OPTIMIZING HAUL ROUTES USING GEOSPATIAL TECHNOLOGIES FOR THE DELIVERY OF READY-MIX CONCRETE IN URBAN AREAS Except where reference is made to the work of others, the work described in this thesis is my own or was done in collaboration with my advisory committee. This thesis does not include propriety or classified information. ___________________________________ Pradeep Suryanarayana Barimar Rao Certificate of Approval: ____________________________ _____________________________ Mary Stroup-Gardiner Wesley C. Zech, Chair Professor Assistant Professor Civil Engineering Civil Engineering ____________________________ ____________________________ Rod E. Turochy John P. Fulton Assistant Professor Assistant Professor Civil Engineering Biosystems Engineering _______________________________ George T. Flowers Interim Dean Graduate School OPTIMIZING HAUL ROUTES USING GEOSPATIAL TECHNOLOGIES FOR THE DELIVERY OF READY-MIX CONCRETE IN URBAN AREAS Pradeep Suryanarayana Barimar Rao A Thesis Submitted to the Graduate Faculty of Auburn University in Partial Fulfillment of the Requirements for the Degree of Master of Science Auburn, Alabama May 10, 2007 iii OPTIMIZING HAUL ROUTES USING GEOSPATIAL TECHNOLOGIES FOR THE DELIVERY OF READY-MIX CONCRETE IN URBAN AREAS Pradeep Suryanarayana Barimar Rao Permission is granted to Auburn University to make copies of this thesis at its discretion, upon request of individuals or institutions and at their expense. The author reserves all publication rights. _____________________________ Signature of Author _____________________________ Date of Graduation iv VITA Pradeep Suryanarayana Barimar Rao, son of Sri. Suryanarayana B. Rao and Smt. Chandravathi B. Rao, was born on October 26, 1977, in Karnataka, India. He completed his diploma in Civil Engineering from C. P. C. Polytechnic, Mysore in 1998. Then he completed his Bachelor of Engineering degree in civil engineering with distinction from Karnataka Regional Engineering College, Surathkal, India. In January 2005, he entered the Graduate School at Auburn University to pursue a Master of Science degree in Transportation Engineering. v THESIS ABSTRACT OPTIMIZING HAUL ROUTES USING GEOSPATIAL TECHNOLOGIES FOR THE DELIVERY OF READY-MIX CONCRETE IN URBAN AREAS Pradeep Suryanarayana Barimar Rao Master of Science, May 10, 2007 (B.E., Karnataka Regional Engineering College Surathkal, India, 2002) 141 Typed Pages Directed by Wesley C. Zech In the ready-mix concrete industry, timely and uninterrupted delivery of ready- mix concrete to a construction site is of utmost importance for efficient construction operation. Due to ever soaring fuel prices and small time windows available for the transport of concrete material (i.e. dependent on the concrete setting time) delivering the material using the shortest distance and shortest time is imperative to register profit. For ready-mix concrete batch plant operations to efficiently schedule ready-mix deliveries, reliable assessment of haul travel time from the batch plant to the construction site is critical. However, the dynamic nature of the roadway network owing to constantly changing traffic conditions makes prediction of travel time complicated. Further, the selection of the optimized route can be difficult to identify by utilizing individual cognitive vi ability. Hence, modern geospatial technologies such as geographic information systems (GIS) and global positioning systems (GPS) are helpful in finding the optimal haul routes. The GIS roadway network and associated traffic data were collected to model the actual roadway network conditions. The GIS roadway network data coupled with traffic information was valuable in identifying optimal haul routes. Time-dependent GIS models were developed to portray the actual traffic conditions on the roadway network for a particular moment in time to calculate a reliable estimation of travel time. GPS data of actual truck haul routes was collected to validate the developed time-dependent GIS models. The time-dependent GIS models developed for actual haul routes exhibited an R2 value of 0.9 in terms of predictive capabilities in determining the actual travel time on the roadway network. The shortest time routes and the shortest distance routes were found for each actual haul route using the developed time-dependent GIS models. The shortest distance models offered a 6.2% savings in distance and shortest time models presented a 16.9% savings in time for haul routes. The savings in haul time will result in higher daily production of the batch plant operation time. The savings in traveled distance could have substantial monetary savings in the long term by reducing the fuel cost associated with the haul operation. vii ACKNOWLEDGEMENTS The author would like to thank his advisor Dr. Wesley C. Zech who has helped the author immensely. The author would like to extend a special thanks to Dr. Zech for his devotion, guidance, and mentorship throughout all phases of this research. The author sincerely thanks Dr. Zech for his support and changing his life by introducing him to a new level of excellence. The author is thankful to Dr. Puneet Srivastava for all his time and guidance of Geographic Information Systems. The author is also thankful to Dr. Robert L. Vecellio, Dr. Brian L. Bowman, and Dr. Rod E. Turochy for their guidance and information on traffic related issues. The author would also like to thank Mr. Charles Bell of Twin City Concrete Co. for permitting the research team to mount Geographic Position systems (GPS) data loggers onto ready-mix concrete trucks and to Dr. John Fulton for providing the GPS data logger and his assistance. The author would also like to thank City of Auburn engineer, Liesa Simpson, for providing traffic and geographic information systems data of the City of Auburn. However, none of this research would have been possible without the love and support of the author?s family and friends: Sri. Suryanarayana Rao, Smt. Chandravathi Rao, Gautham, Prathap, Sandeep, Anjani, and Suresh. viii Style manual or journal used Auburn University Graduate School Guide to Preparation of Master?s Thesis Computer software used Microsoft Word, Microsoft Excel, Paint, Adobe Acrobat 6.0, and ArcGIS 9.1 ix TABLE OF CONTENTS LIST OF TABLES............................................................................................................. xi LIST OF FIGURES ......................................................................................................... xiii CHAPTER ONE..................................................................................................................1 INTRODUCTION ...............................................................................................................1 1.1 Background........................................................................................................1 1.1.1 Cycle Time ..........................................................................................1 1.1.2 Importance of Travel Time Reliability................................................4 1.1.3 Traffic Congestion...............................................................................4 1.1.4 Geospatial Technology ........................................................................8 1.1.5 Importance of Reduced Haul Time......................................................8 1.2 Research Objectives.........................................................................................10 1.3 Organization of Thesis.....................................................................................12 CHAPTER TWO...............................................................................................................14 LITERATURE REVIEW ..................................................................................................14 2.1 Introduction .....................................................................................................14 2.1.1 Modeling for Efficient Route Finding ...............................................16 2.2 Component Properties of Roadway Network..................................................18 2.3 Time to Traverse a Street Segment..................................................................21 2.4 Delay Function For Signalized Intersection ....................................................22 2.5 Delay Function For UnSignalized Intersections..............................................28 2.6 Geographic Information Systems (GIS) ..........................................................30 2.6.1 Map Projections and Coordinate Systems .........................................36 2.6.2 Digital Orthophoto Quad (DOQ).......................................................43 2.7 ArcGISTM software ..........................................................................................43 2.8 Global Positioning Systems.............................................................................45 2.8.1 How GPS Works ...............................................................................46 2.9 Summary..........................................................................................................48 CHAPTER THREE ...........................................................................................................49 DATA COLLECTION ......................................................................................................49 3.1 Collection of GIS Data ....................................................................................49 3.1.1 GIS Streets Data for the City of Auburn ...........................................50 3.1.2 GIS City of Auburn Traffic Data.......................................................51 3.1.3 COA GIS Traffic Signal Data ...........................................................53 3.2 Incorporation Of Critical COA GIS Roadway Network Parameters...............54 x 3.2.1 Number of Lanes ...............................................................................54 3.2.2 Traffic Signal Locations ....................................................................55 3.3 Traffic Data Collection....................................................................................57 3.3.1 Annual Average Daily Traffic (AADT) Values ................................57 3.3.2 Traffic Multiplication Factors............................................................58 3.3.3 Percentage of Traffic Growth ............................................................60 3.3.4 GPS Data Logger...............................................................................61 3.4 GPS Data Collection for Concrete Truck Haul Route.....................................62 3.4.1 GPS Data Collection..........................................................................66 3.4.2 Collected GPS Data ...........................................................................68 3.5 Conclusion .......................................................................................................70 CHAPTER FOUR .............................................................................................................72 GIS MODEL DEVELOPMENT, VALIDATION, AND OPTIMIZATION....................72 4.1 Introduction .....................................................................................................72 4.2 Base GIS Model Development Methodology .................................................74 4.2.1 Signalized Intersection Delay............................................................74 4.2.2 Unsignalized Intersection Delay........................................................78 4.2.3 Updating Speed Limit Data ...............................................................79 4.2.4 Traffic Growth Factors ......................................................................80 4.2.5 Network Travel Time ........................................................................82 4.3 Time-Dependent GIS Models..........................................................................85 4.3.1 Calculating Signalized Intersection Delay.........................................89 4.3.2 Calculating Network Travel Time.....................................................90 4.4 Time-Dependent GIS Model Validation .........................................................91 4.5 Optimization Of Haul Routes ..........................................................................97 4.6 Conclusion .....................................................................................................110 CHAPTER FIVE .............................................................................................................111 CONCLUSIONS AND RECOMMENDATIONS..........................................................111 5.1 Contributions Of The Research .....................................................................111 5.2 Suggestions for Future Research ...................................................................113 REFERENCES ................................................................................................................116 APPENDIX A..................................................................................................................120 xi LIST OF TABLES Table 2.1 Definition of Signalized Intersection Terms ................................................... 23 Table 2.2 GIS Software Producers and Software Products. ............................................ 35 Table 3.1 Description of Shape File GIS Data????????????????..49 Table 3.2 Description of Relevant Field Contained in StreetCL Attribute Table ........... 51 Table 3.3 Description of Each Field in AADT2006 attribute table................................. 52 Table 3.4 Description of Each Field in Signals Attribute Table...................................... 54 Table 3.5 Hourly Variations of Traffic Volumes on the COA Roadway Network......... 59 Table 3.6 Daily Variation of Traffic Volume on the COA.............................................. 59 Table 3.7 Seasonal Variation of Traffic per Month......................................................... 60 Table 3.8 Column Descriptions of Raw Data.................................................................. 68 Table 4.1 Assumptions for Development of Time-Dependent GIS Models ?????73 Table 4.2 VB Script Used to Update Traffic Signal Cycle Length ................................. 76 Table 4.3 VB Script of for Weight of The Link .............................................................. 77 Table 4.4 VB Script for Speed Limit Update .................................................................. 80 Table 4.5 VB Script for Projecting the AADT Data ....................................................... 82 Table 4.6 VB Script for furnishing Capacity of Roadway .............................................. 84 Table 4.7 VB Script for Getting Length of the Roadway Link ....................................... 84 Table 4.8 Date and Time Information Used for GIS Models.......................................... 86 xii Table 4.9 VB Script for Calculating Delay at Signalized Intersection............................ 89 Table 4.10 VB Script for Travel Time Calculation ......................................................... 90 Table 4.11 Data Summary for GIS Model Validation..................................................... 93 Table 4.12 Comparison of Actual Route and Modeled Routes..................................... 101 Table 4.13 Alternative GIS Modeled Shortest Distance Routes Compared Against Actual Routes Taken............................................................................................... 105 Table 4.14 Alternative GIS Modeled Shortest Time Routes Compared Against Actual Routes Taken .......................................................................................................... 107 Table 4.15 Savings Experienced by GIS Model Models.............................................. 109 xiii LIST OF FIGURES Figure 1.1 Flow Chart of Ready-Mix Concrete Truck Cycle Time Operation. ................ 2 Figure 1.2 Weekday Peak-Period Congestion Growth Over the Past 20 Years in the...... 6 Figure 1.3 Sources of Congestion...................................................................................... 7 Figure 2.1 Definition of Total, Stopped, Deceleration, and Acceleration Delays........... 24 Figure 2.2 GIS Map Layer System.................................................................................. 32 Figure 2.3 Earth Divided into Latitude and Longitude. .................................................. 36 Figure 2.4 Measurement of an Earth Feature. ................................................................. 38 Figure 2.5 Model of the Earth.......................................................................................... 39 Figure 2.6 Earth Surfaces Based on GIS. ........................................................................ 40 Figure 2.7 Geometric Representations of Projections. .................................................... 41 Figure 2.8 Zones of State Plane Coordinate (SPC) System in the USA.......................... 42 Figure 2.9 Constellation of GPS Satellites Around the Earth. ........................................ 46 Figure 2.10 Time Difference Between Radio Waves...................................................... 47 Figure 3.1 City of Auburn GIS Data of the Roadway Network. ..................................... 50 Figure 3.2 City of Auburn AADT Data........................................................................... 52 Figure 3.3 City of Auburn Traffic Signal Data. .............................................................. 53 Figure 3.4 Number of Lanes Incorporated in the COA Roadway Network GIS Data.... 56 Figure 3.5 Traffic Signal locations Incorporated into the COA Roadway Network....... 56 xiv Figure 3.6 AADT values Incorporated Into the COA Roadway Network. ..................... 57 Figure 3.7 GPS Data Logger. .......................................................................................... 62 Figure 3.8 GPS Data Logger Compact Flash Drive Mounting Area............................... 62 Figure 3.9 Ready-mix Concrete Truck. ........................................................................... 64 Figure 3.10 GPS Receivers Mounted on Ready-mix Concrete Trucks........................... 65 Figure 3.11 GPS Instrument Mounted Inside the Truck.................................................. 65 Figure 3.12 GIS Road Network at Batch Plant Location. ............................................... 67 Figure 3.13 GPS Data Stored on the GPS Data Logger Instrument Flash Card. ............ 68 Figure 3.14 Representative Haul Route GPS points on GIS Road Network................... 70 Figure 4.1 Attribute Data Till Projection of AADT ......................................................... 76 Figure 4.2 Similarity of GIS Model of Actual Haul Route and GPS Haul Routes. ........ 92 Figure 4.3 GIS Model Validation for Actual vs. Predicted Travel Time of Haul Routes. .................................................................................................................................. 96 Figure 4.4 GIS Model Validation for Actual vs. Predicted Travel Time of Return Routes. .................................................................................................................................. 97 Figure 4.5 Actual and Modeled GIS Routes.................................................................. 100 1 CHAPTER ONE INTRODUCTION 1 1.1 BACKGROUND Mobility has always been important to human society. Transportation is the movement of people and goods and an efficient transportation system is essential for the economic improvement of any region. The speed, cost, and capacity of available transportation has a significant impact on the economic health of an area. For any trade, cost and time associated with the transportation of raw and finished goods is essential for registering profit. The same applies to the concrete construction industry. Optimizing the time and cost associated with delivering ready-mix concrete from the batch plant to the job site is one key factor to registering profit during a ready-mix plant operation. If the transportation cost associated with ready-mix concrete delivery is substantial, and if this process is not optimized, the batch plant may loose its competitive edge over other suppliers. 1.1.1 Cycle Time Cycle time is the total time it takes to complete the sequence of an operation that is repeated regularly from beginning to the end. For contractors and ready-mix concrete batch plant companies to efficiently schedule ready-mix deliveries, cycle time is 2 critical. Figure 1.1 shows the cycle time of a ready-mix concrete truck operation. Figure 1.1 Flow Chart of Ready-Mix Concrete Truck Cycle Time Operation. The loading operation of a ready-mix concrete truck regarding cycle time includes various operations. Once the order to dispatch the ready-mix concrete is received, total loading time is comprised of time spent under the silo to receive the ready-mix, rinsing of the concrete truck, filling the water tank of the truck, and obtaining the delivery ticket. The time required to haul the ready-mix from the batch plant to the job site is considered in the haul operation. Discharging is considered when the concrete is being poured to its final location on the job site. After the pouring of concrete is complete, the truck has to be cleaned before it returns to the batch plant. Wash-out is the operation of washing the concrete truck and chute on the job-site. The last component of the cycle time operation is the return haul of the truck back to the ready-mix batch plant. Of all these factors, the haul time to the job site and the return haul time to the plant are the only elements not under the full control of the concrete batch plant. The reasoning 1. Load 2. Haul 3. Discharge 4. Wash-out 5. Return 3 is because travel times between two locations will vary depending on traffic congestion, roadway under construction, vehicular accidents, and local weather. The aforementioned conditions may cause the traffic patterns to fluctuate heavily, which may result in haul operations taking twice the time in abnormal traffic conditions (i.e. traffic congestion) compared to the haul time experienced during normal traffic conditions. Of all these, traffic congestion is the primary cause for delay. If the batch plant can accurately predict haul times, they will have the ability to reduce the waiting time trucks may experience as result of queuing on the job site if multiple trucks are waiting to discharge the ready-mix concrete. This will enable the batch plant to reduce the total cycle time and associated cost of operation. Predicting haul times accurately will allow batch plants to effectively plan and schedule their hauling operation. In the case of a concrete haul truck operation the contractor must plan and prepare for the arrival of the concrete material. When a contractor requests a concrete delivery, he needs to be ready to place that concrete once it arrives. On-time delivery of concrete is essential to both the contractor and concrete batch plant. If a truck arrives early, the concrete placement crew may not be prepared to discharge the material due to formwork or rebar not being completely placed. In this situation trucks are expected to spend additional time waiting on-site prior to discharge. This results in a longer than anticipated cycle time and causes additional cost to the ready-mix plant. Even worse if the concrete sets too much, the window for placement is lost and the entire load may have to be disposed of. If the concrete truck arrives late, the contractor?s crew may be idle because the material delivery is the controlling activity. 4 1.1.2 Importance of Travel Time Reliability According to Batley, a ?reliability premium? measure is the delay in arrival time that an individual would be willing-to-pay in exchange for eliminating unreliability in arrival time for a given departure time [Batley, 2006]. Prevalence and magnitude of cost for an individual towards reliable travel time is different. According to a study conducted by Lam and Small, the value of time for commuters is found to be $22.87 per hour [Lam and Small, 2001]. In the same fashion the cost reliability in travel time for a haul truck may be as much as wages of a whole concrete placement crew at the job site for the period that the truck arrived late. As the crew size and wages varies per job site, the cost of reliability of concrete trucks may also vary. 1.1.3 Traffic Congestion Traffic congestion is the primary factor resulting in unreliable travel times. Traffic congestion is defined as an excess of vehicles on a portion of roadway at a particular time, resulting in less than normal speeds (i.e. free-flow speeds). Traffic congestion occurs when the volume of traffic on a roadway is high enough to become detrimental to its performance vis-?-vis reductions in vehicle speeds, increases in travel time, and increases in fuel consumption. Severe congestion occurs when vehicles on a roadway experience stop-and-go conditions or stopped traffic. One significant element of congestion is the cost of additional time and wasted fuel [Schrank and Lomax, 2005]. Congestion is measured based on travel time experienced by users of the highway system and wasted fuel. According to the 2005 Urban Mobility Report published by the 5 Texas Transportation Institute (TTI), in 2003, the 85 largest metropolitan areas experienced 3.7 billion vehicle-hours of delay, resulting in 2.3 billion gallons in wasted fuel and $63 billion in lost productivity [Schrank and Lomax, 2005]. Congestion continues to grow, and there is evidence that user trips are becoming increasingly unreliable [Margiotta and Taylor, 2006]. A study conducted by the Texas Transportation Institute (TTI) indicated that in 2003 it took 37% longer, on average, to make a peak period trip in urban areas compared with the time it would take if traffic flowed freely [USDOT, 2005]. The development and monitoring of reliability measures such as extent, duration, and intensity is a critical first step in understanding the complete congestion picture [Margiotta and Taylor, 2006]. Extent of congestion is measured by the percentage of time the travel was congested. Duration of congestion is measured by the number of hours in a day traffic was congested, while congestion intensity is measured as a percent increase in average delay. Figure 1.2 shows how week day peak period congestion has drastically increased in several ways over the past 20 years in the largest U.S. cities. Figure 1.2 illustrates that over the past 20 years, the congestion has grown in intensity, extent, and duration. From 1982 to 2003 the duration of the daily congestion period has increased from 4.5 hours to 7 hours per day. During that period the extent of congestion has doubled from 33% to 67% of travel time. In the same period, the intensity of congestion has tripled from 13% to 37% on average per day. Lastly, its been seen that now congestion affects more periods of the day, not just rush hour or peak hour periods. This unpredictable variability in travel time has led to the decrease in travel time reliability. Haul truck operations need to 6 consider placing a buffer time into their haul trip to account for this travel time variability. Figure 1.2 Weekday Peak-Period Congestion Growth Over the Past 20 Years in the Largest U.S. Cities. Source: [Margiotta and Taylor, 2006] According to the Federal Highway Administration (FHWA) congestion occurs when the free flow of traffic on a roadway is impeded due to excess vehicle demand, construction and maintenance activities, traffic incidents, inclement weather, or other road conditions and events [FHWA, 2006b]. Figure 1.3 shows the percentage of congestion each congestion source contributes. According to FHWA 40% of the congestion is caused by insufficient capacity (i.e., bottlenecks), 25% by incidents (i.e., crashes, disabled vehicles, 15% by weather (i.e., snow, ice, and fog), 10% by work zones, 5% other non- recurring events, and 5% by poor signal timing. The major source of congestion is insufficient roadway capacity or a bottleneck situation. A bottleneck occurs when a particular section of roadway traffic increases beyond its designed capacity. A traffic Extent Congestion In 1982 Intensity 13 % average delay 33 % of travel Duration 4.5 hrs / day Congestion In 2003 7 hrs / day Extent 67 % of travel 37 % average delay Intensity Duration 7 incident is an emergency road user occurrence, a natural disaster, or an other unplanned event that affects or impedes the normal flow of traffic [FHWA, 2003]. Debris in travel lanes, vehicular crashes, natural disasters, and breakdowns are all common type of incidents that are unplanned events and effect or impede the normal flow of traffic. A work zone is an area of a highway experiencing construction, maintenance, or utility work activities [FHWA, 2003]. A work zone may result in physical changes to a highway environment that may include: a reduction or elimination of shoulders, reduction in the number or width of travel lanes, lane shifts, and lane diversions. Bad weather (i.e., snow, ice, and fog) may change the characteristics of driver behavior and vehicle stability leading to reduction in vehicle speeds. Poor signal timings will force travelers on roadway networks to wait unnecessarily at traffic signals. A special event may bring a sudden influx of traffic onto a roadway network for which it was not designed to accommodate. Figure 1.3 Sources of Congestion. Source: [FHWA, 2006b ] Weather 15% Incidents 25% Work zones 10% Insufficient Capacity 40% Other non-recurring events 5% Poor signal timing 5% 8 1.1.4 Geospatial Technology Geospatial Technologies can be categorized into two categories: (i) Geographic Information Systems (GIS) and (ii) Global Positioning Systems (GPS). GIS aids in visualizing the roadway network by mapping data spatially. GIS allows the users to view the entire network, the grades of the roads, width of the roadways, speed limits, intersection signal locations, and various other existing features of the roadway network. GPS can be used to locate and spatially track a haul truck on a roadway network in real-time. GIS and GPS technologies combined can be utilized to track and analyze haul truck movement on the roadway network in real-time. These two systems used in combination can accurately provide assurance regarding the route the haul trucks have traveled, and also helps to verify that the haul was completed within a specified time period. However the information provided by the combination of these systems is inadequate to furnish the time required for the trucks to travel from their origin to their destination in a dynamic roadway network. GPS units do not furnish the roadway network attributes such as traffic information of the roadway which are dynamic. 1.1.5 Importance of Reduced Haul Time The time window between adding water to concrete at the batch plant and pouring the concrete at the job site is critical. This time window for the concrete in a revolving rotating concrete truck drum should be less than the time required for the concrete to begin setting. Initial time of set is the elapsed time after initial contact of cement and 9 water required for sieved mortar to reach a penetration resistance of 3.5 MPa (500 psi). This definition is according to ASTM C 403 and AASHTO T 197 [TXDOT, 2006]. The time required to travel from the batch plant to a job site is a major consumer of this critical time. Reducing haul time will increase the total supply capacity of a ready-mix concrete plant. Due to escalated fuel prices, fuels currently impact the total operating cost of haul trucks. On 20th May 2006 according to Mr. Charles Bell, the head of operations at the concrete batch plant in Auburn, AL, fuel is consumed at a rate of 1 gallon for 2.1 miles of travel. Due to this high truck operating cost, it will be beneficial to find the cheapest routes between the batch plant to construction sites for haul truck operation. The cheapest route may either be the shortest distance route or the quickest route. Internet tools, such as Google Map and Mapquest, can provide haul route information based on minimum travel time or distance. These tools do not consider the variability and dynamics of roadway network such as the traffic condition at that time of day when calculating travel times. This is evident from the fact that travel time estimation provided by these internet tools for a particular route does not change when it is obtained during different times of the day. Hypothetically, a haul route with longer downhill and shorter uphill journeys will be more economical than longer uphill and shorter downhill journeys, therefore resulting in a more profitable operation. Fuel cost primarily depends on the length of haul route and 10 the number of stops or idle times encountered traveling the route, such as traffic signals or congested traffic. All the above factors enforce the demand for finding a method for calculating the cheapest route at a given time of day and for accurately forecasting travel time. 1.2 RESEARCH OBJECTIVES The primary goal of this research is to develop a GIS based model to assist in determining optimal routes from one location to other in urban areas to transport material in the most economical and efficient manner. The aim is to make the haul and return travel times more reliable during the decision making process when scheduling trucks. Traffic demands tend to vary significantly depending on the season of the year, the day of the week, and even the time of day [FHWA, 2006b]. The major differences in travel time occur mainly due to recurring instances and variations in levels of traffic congestion throughout the year [Eglese et al., 2006]. Therefore historical traffic volumes for a given hour of the day will be an important parameter. Secondly, this traffic volume data can be modified to adapt to the variation in traffic volume using typical hourly, daily, and seasonal traffic factors. Traffic factors are provided by the Institute of Transportation Engineers? Traffic Engineering Handbook or derived for an area by using historical traffic count of the area. Eglese et al., also found that time- dependent vehicle routing and scheduling systems are benefited from using real-word data for the road network [Eglese et al., 2006]. If traffic demand and other physical characteristics of roadways and work zone activity are known, then the time of congestion occurrence due to insufficient capacity and work zones can be forecasted. 11 Similarly, the time required to travel at a particular hour of day on a roadway and through a work zones can also be estimated. In the same fashion, intersection delay functions can be used to determine additional time required to traverse an intersection. By incorporating all these above methods, a reliable travel time through a dynamic roadway network can be computed. The optimal route will be found using GIS software for determining the shortest and/or fastest route between a specified origin and destinations in an urbanized area while considering the traffic conditions for a particular time period. The software will allow us to calculate the required travel time and distance for each haul route. For the GIS software to understand and analyze the functionality of the roadway network, information regarding the highway network, such as: roadway interconnectivity, length, speed limits, number of lanes, grades, weight limits and restrictions, annual average daily traffic (AADT), location of traffic signals, and intersection signal timing is required. The location of both the batch plant and the project will be necessary to determine which route within the network the hauling operation will be traveling. The information pertaining to the characteristics of the construction operations (i.e., as the scheduling aspects) and the truck performance characteristics (i.e., weight, capacity, and fuel efficiency) will also be required. With the aid of all the above conditions, optimal routes for each truck at a given hour of a specific day hauling material to a project location will be determined using GIS software. The various types of data required and collected to find the optimal route will aid in developing a GIS database for the roadway network under investigation. During the project execution, GPS instruments 12 will be mounted on two hauling trucks to measure the accuracy of the optimal route found using GIS software. Based on the collected GPS data, necessary adjustments to the GIS model will be made to refine the method for calculating the optimal route. The specific objectives of the research are as follows: 1. Collect various GIS and traffic data related to the City of Auburn (COA) roadway system for the development a base GIS roadway network model. 2. Collect GPS data of a trucking operation to obtain actual spatial movement and travel time information for the routes traveled by the haul trucks. 3. Develop time-dependent GIS roadway network models to accurately predict truck travel time and compare the results with the actual travel times obtained through GPS data collection. 4. Find the shortest time route and shortest distance route for haul routes using the GIS time-dependent models and optimization techniques. 1.3 ORGANIZATION OF THESIS This thesis is divided into six chapters that clearly organize, illustrate, and describe the steps taken to meet the defined research objectives throughout the duration of this project. Immediately following this chapter, Chapter 2: Literature Review, summarizes the body of knowledge pertaining to this study and synthesizes previous research efforts. The focus of the literature review centered upon the travel time required to traverse urban roadways and intersections. Chapter 3: Data Collection, describes the capabilities and requirements of the GIS and GPS applications. Chapter 3 also explains the process of collecting relevant traffic data and GIS information from 13 the Internet, or public and private institutions for the development of the City of Auburn (COA) GIS roadway network. A description of the GPS equipment utilized for the data collection of ready-mix concrete hauling operations is also provided. Chapter 4: GIS Model Development, Validation, and Optimization, explains the process followed to develop a roadway network in GIS. Chapter 4 further explains the procedures followed and the use of the GIS software?s Network Analyst tool to find the shortest and cheapest haul route during a certain time of the day. It describes the process followed to validate the developed GIS models based on the field GPS data collected. Finally, Chapter 5: Conclusions and Recommendations, provides input regarding adequacy of model developed, while identifying the potential for further research that can be conducted to improve upon this research effort. 14 CHAPTER TWO LITERATURE REVIEW 2 2.1 INTRODUCTION Locomotion is the physical movement through space, perceiving objects and the environment they are in, and moving towards a visible place or object while avoiding various obstacles [S. M. Freundschuh, 2000]. Wayfinding is the thought process and thinking of how the movement to the distant location or the location that cannot be seen from the current position in the environment [S. M. Freundschuh, 2000]. Selecting routes for travel, planning trips, and making estimations of distance, and travel time are some of the wayfinding behaviors. The commute, the explore, and the quest are three categories of wayfinding tasks [S. M. Freundschuh, 2000]. Traveling along a familiar route between two known places is called commute. The success of this form of wayfinding is measured in terms of travel time (i.e., shortest time), the amount of effort to negotiate the route (i.e., the fewest stops and turns), and arriving at the destination is considered as a certain event. Home to college trip, office to bank trip, home to store trip of a person are some typical commute types of wayfinding. The second wayfinding task type is explore. An explore type of wayfinding is when start and end points of a trip are known and travel is through an unfamiliar area. If a person travels from his house to a bank on a new route 15 than their usual route may be labeled as explore. Explore requires strategic planning of the routes to travel and travel directions. Quest is third type of wayfinding. A quest begins at a known place and ends at a unknown place where navigation is similar to explore type. Out of these three wayfinding tasks, this research work is focused on the mixture of commute and explore. During concrete batch plant operations the fleet of trucks are usually operated by drivers who have fair knowledge about travel routes and places in the marketing (i.e., supply) area of the batch plant. Even the new drivers become well versed with the travel route area in a short period of time. Inexperienced truck drivers typically follow the explore method of wayfinding and apply their cognitive knowledge to travel from the batch plant to the concrete dispatch location. When the truck drivers receive a concrete dispatch order to a similarly traveled or nearby location they tend to follow previously explored paths. Truck drivers perceive these paths as the shortest distance path or shortest time path from batch plant to dispatch location based upon field experience. But often the actual shortest distance or shortest time path is different than the route selected by the truck drivers. In urban areas the amount of traffic on a roadway varies considerably over the course of a day, week, and month. The traffic intensity on a roadway is dependent on various factor mentioned earlier (i.e., when the travel takes place, the amount of construction work the network is experiencing, the number of accidents, and local weather). Therefore the fastest route from an origin to a destination may not be same during morning and afternoon periods of time. With the help of GIS tools and its spatial representation of the roadway network, it is possible to 16 identify the true shortest distance and time routes within a network. Users can continuously update the GIS roadway network with current traffic information. It is possible to incorporate known variations of traffic into the GIS roadway network to find true shortest distance and time routes for a given period of time. 2.1.1 Modeling for Efficient Route Finding In the ready-mix concrete delivery industry, efficiently routing and scheduling of a fleet of trucks to service customers plays an important role. However routing and scheduling is difficult to manage due to the fact that this widely studied research domain lacks the modeling approach that more closely represent real-life conditions [Ichoua, 2003]. In vehicle routing models, travel times or variation in travel times on the roadway network is sparsely addressed due to the dynamic nature of roadway network. Achieving efficient vehicle routing models will depend on good trade-offs between implementation requirements and the ability to reflect the complexity of real- world conditions such as fluctuations in travel times [Ichoua, 2003]. Tarantilis and Kiranoudis (2002) have explored a spatial decision support system for solving the vehicle routing problem. The main focus of their project was to deliver the goods to all customers using minimum number of vehicles in the fleet while traveling a lesser distance. A secondary focus of their research was to minimize the total distance traveled by all the vehicles. The shortest path or route for each vehicle was found in a specified road network using the network analysis tool of GIS software. The network analysis tool in turn used Dijkstra?s algorithm to find the shortest distance. In their 17 project, time required to traverse the route was not considered [Tarantilis and Kiranoudis, 2002]. The shortest distance route may take more time in the dynamic nature of the roadway network. In their research vehicle routing techniques are based only on distance which may cause delay in delivering the goods to the customer. Time-dependent vehicle routing is very difficult to model and rarely addressed. But time-dependency is widely studied in time-dependent traveling salesman problems and shortest path problems. In some of earlier research work on vehicle routing, fixed travel time for the entire network was considered [Ichoua et al. 2003]. Ichoua et. al. showed that a vehicle routing time-dependent model provides better results over a vehicle routing based on fixed travel times [Ichoua et al. 2003]. The authors incorporate time- dependency of travel time in vehicle routing models by taking time-dependent travel speeds into account. They have also considered the variability of travel speeds on different network links or streets. In a GIS roadway network a street segment is commonly referred to as a link and an intersection is commonly known as a node. Ichoua et. al. have explored to model travel time for a set of fixed movements or links with the time-dependent travel speed model using a parallel tabu search. Their model is not directly applicable for practitioners since the travel route found for a vehicle at particular dispatch time using a parallel tabu search is difficult to visualize. Moreover they do not consider the delay experienced by the vehicles caused at every intersection of a roadway network. When calculating the total travel time in a network, not only the variation in travel speeds on a street but the delay experienced at the intersection should also be considered. 18 The time-dependent vehicle routing explored by Fleischmann et. al. and Kim et. al. based on online or real-time traffic information is not feasible for every geographic area [Fleischmann et al., 2004 and Kim et al., 2005]. The reason being that each of their models require extensive infrastructure by means of Intelligent Transportation System (ITS) facilities in the area to obtain real-time traffic information or travel time estimations from a Traffic Management Center (TMC). Another limitation of these models is that an ITS and TMC facility in an area of a roadway network may not necessarily provide travel time or traffic information for each roadway link. In the U.S. ITS is deployed in 108 metropolitan areas out of a total of 280 metropolitan areas [U S Censes Bureau, 2006 and USDOT, 2006]. 2.2 COMPONENT PROPERTIES OF ROADWAY NETWORK According to the American Association of State Highway and Transportation Officials (AASHTO) there are four functional highway systems for urban areas used to classify roadways that include: (i) urban principal arterials (streets), (ii) minor arterials (streets), (iii) collector streets, and (iv) local streets. Arterial streets are roads that primarily serve longer trips. Arterials also provide access to abutting commercial and residential land uses. The urban principal arterial system constitutes a small percentage of the total roadway network while serving a high proportion of total urban travel. The principal arterial system carries most of the through movements bypassing the Central Business Districts (CBD) and most of the trips entering and leaving the urban area AASTHO, 2004]. It also serves significant intra-area travel. Intra-area travel includes 19 travel between CBD and outlying residential areas as well as travel between major inner-city communities. It also provides travel between major suburban centers and continuity for all rural arterials that intercept the urban boundary. The arterial street augmented and interconnected with the urban principal arterial system is called the urban minor arterial system. The urban minor arterial system offers lower traffic mobility and stresses importance on land access more so than the major arterial system. Further land access and traffic circulation within residential, commercial, and industrial areas are provided by collector streets. The access function of collector streets is more important than that of arterials. The collector system may penetrate residential neighborhoods and distribute trips from the arterials to their ultimate destinations. Also collector streets collect traffic from local streets in residential neighborhoods and channel it onto the arterial streets. The remaining roads of a street system facility not categorized as an arterial or collector are considered local streets. Local streets provide connections to higher order street systems and allocate direct access to abutting lands. In the hierarchy of urban street transportation facilities, urban streets (including arterials and collectors) are ranked between local streets and multilane suburban and rural highways, according to Highway Capacity Manual [TRB, 2000]. Downtown streets are signalized street facilities that often resemble arterials. The function of downtown streets can change with the time of day [TRB, 2000]. ?Multilane rural highways differ from urban streets in the following ways: roadside development is not as intense, density if the traffic access points is not as high, and signalized intersections are 2 miles apart? [TRB, 2000]. 20 Street environments, interaction among vehicles, and traffic control are three main factors influencing the speed of vehicles on urban streets [TRB, 2000]. Spacing between signalized intersections, the number and width of lanes, speed limit, types of median, driveway/access point density, level of pedestrian activity, and existence of parking, comprise the street environment of the urban street facility. Traffic density, turning movements, and the proportion of trucks and buses provide additional detail about the interaction among the vehicles. Traffic controls (i.e., including signals and signs) force a portion of all vehicles to slow or stop. ?Free flow speed (FFS) refers to the speed chosen by the average driver when vehicle interaction and traffic control are not factors. FFS is the average speed of the traffic stream when traffic volumes are sufficiently low that drivers are not influenced by the presence of the other vehicles and when intersection traffic control (i.e., signal and sign) is not present or sufficiently distant as to have no effect on speed choice? [TRB, 2000]. The presence of other vehicles restricts the speed of a vehicle in motion because of differences in speeds among drivers or because downstream vehicles are accelerating from a stopped condition and have not yet reached FFS. These conditions are recurring over the course of a day causing a driver not to travel at the FFS. The running speed is computed as the length of the segment of street divided by the average running time [TRB, 2000]. ?The running time is the time taken to traverse the street segment, less any stop-time delay? [TRB, 2000]. The presence of traffic control device on an urban street segment tends to reduce vehicle speeds below the average running speed. The average travel speed captures the effect of traffic control, and this 21 speed is computed as the length of the segment divided by the average travel time. The travel time is the time taken to traverse the street segment, including any stop-time delay [TRB, 2000]. 2.3 TIME TO TRAVERSE A STREET SEGMENT The time required to travel on network roads (links) and through intersections (nodes) is essential for determining the fastest route. The method used by Hedayat and Iravani, for calculating the travel time on a network link, did not consider the effect that traffic traveling in the opposing direction has on travel time [Hedayat and Iravani, 1999]. Traffic traveling in the opposing direction on the same roadway is important as it affects and lessens the travel speed of the roadway. Hedayat and Iravani used a volume delay function for calculating delay at each network link shown in equation 2.1 below: 4 0( ) * * 1 0.15* * xt x l t c w ? ?? ?= + ? ?? ?? ?? ? ? ? (2.1) where: t(x) = travel time on the link (min), l = length of the link (ft), t0 = travel time for traveling unit length at free flow speed of the link (min), x = traffic on the link (veh/h), c = capacity of the link per lane (veh/h), and w = number of lanes in the link. 22 The capacity of an urban street is primarily related to the signal timings and geometric characteristics of the facility as well as to the composition of traffic on the facility. Geometric characteristics of the facility are fixed. Thus, while traffic composition may vary over time, the capacity of the facility is generally a stable value [TRB, 2000]. According to HCM the minor and major urban arterials have an adjusted saturation flow rate of 1,700 pc/hr/ln [TRB, 2000]. By referring to exhibit number 10-7 of HCM, service volume capacity of urban arterials is 800 veh/hr for one lane roadways, 1,620 veh/hr for two lane roadways, 2,430 veh/hr for three lane roadways, and 3,250 veh/hr for four lane highways. These values are derived by assuming signal density of 10 per mile, FFS of 30 mph, cycle length of 70 seconds, effective green ratio of 0.45, adjusted saturation flow rate of 1700 pc/hr/ln, peak hour factor of 0.92, percentage of left and right turning vehicles as 10%, and the existence of left turn bay. Effective green ratio is the effective green time divided by the signal cycle length. A highway is in a base condition when it has a divided multilane highway, on a level terrain, with 12 ft lane widths and 6 ft shoulder widths and contains only passenger cars. 2.4 DELAY FUNCTION FOR SIGNALIZED INTERSECTION An accurate estimation of intersection delay will lay the foundation for better traffic assignment models. However, any stop-time or intersection delay is a difficult parameter to estimate. Teply (1989) indicated that a field-measured delay and delay obtained from analytical formulas can not be matched accurately [Teply, 1989]. The intersection delay has its direct relation to what motorists experience while attempting 23 to cross an intersection. At signalized intersections delay is computed as the difference between the travel time that is actually experienced by a vehicle while progressing through the intersection and the travel time this vehicle would have experienced in the absence of traffic signal control. Table 2.1 presents a list of HCM definitions for the terms that are used in calculating or representing signalized intersection delay. Table 2.1 Definition of Signalized Intersection Terms Source: Highway Capacity Manual 2000 [TRB, 2000] Name Definition Unit Symbol Control Delay The component of delay that results when a control signal causes a lane group to reduce speed or to stop; it is measured by comparison with the uncontrolled condition Sec d Cycle A complete sequence of signal indications Cycle length The total time for a signal to complete one cycle Sec C Clearance lost time The time between signal phases during which an intersection is not used by any traffic Sec l2 Effective green time The time during which the given traffic movement or set of movements may proceed; it is equal to the cycle length minus the effective red time Sec gi Effective red time The time during which a given traffic movement or set of movements is directed to stop; it is equal to the cycle length minus the effective green time Sec ri Extension of effective green time The amount of the change and clearance interval at the end of the phase for a lane group, that is usable for movement of its vehicles Sec e Green time The duration of the green indication for a given movement at a signalized intersection Sec Gi Interval A period of time in which all traffic signal indications remain constant Lost time The time during which an intersection is not used effectively by any movement; it is the sum of clearance lost time plus start-up lost time Sec tL Phase The part of the signal cycle allocated to any combination of traffic movements receiving the right-of-way simultaneously during one or more intervals Red time The period in the signal cycle during which, for a given phase or lane group, the signal is red Sec Ri Saturation flow rate The equivalent hourly rate at which previously queued vehicles can traverse an intersection approach under prevailing conditions, assuming that the green signal is available at all the times and no lost times are experienced Veh/h si Start-up lost time The additional time consumed by the first few vehicles in a queue at a signalized intersection above and beyond the saturation headway, because of the need to react to the initiation of the green phase and to accelerate Sec l1 24 Figure 2.1 is a diagram used to illustrate the total delay experienced by a vehicle at a signalized intersection. The total delay is the sum of deceleration delay, stopped delay, and start-up lost time (i.e., acceleration delay). The figure shows the vehicle path and time the vehicle may have taken to traverse the specified distance. The figure also shows three cases of vehicle paths namely: 1. vehicle path if there is no traffic control, 2. actual vehicle path considering stopped delay, gradual deceleration and acceleration delay, and 3. vehicle path considering stopped delay, instantaneous deceleration and acceleration delay. Typically, transportation professionals define stopped delay as the delay incurred when a vehicle is fully immobilized. The delay incurred by a decelerating or accelerating vehicle is categorized as deceleration and acceleration delay, respectively. Figure 2.1 Definition of Total, Stopped, Deceleration, and Acceleration Delays. Source: [Dion et al., 2004][11] 25 Acceleration delay, a fraction of the total delay, is originated from the time required by individual drivers to react to changes in the signal display at the beginning of the green interval, to mechanical constraints, and to individual driver behavior. A part of the total delay is stopped delay and attributed to traffic signal operation. In the perfect scenario, vehicles coming to a signalized intersection would stop instantaneously on red signal display and vehicles queued at an intersection would start moving at their ideal speed immediately following the display of a green signal. The deterministic queuing model for predicting delay at signalized intersections assumes the number of vehicles that can be served during a green signal interval is greater than the number of arrivals per cycle [Dion et al., 2004]. The deterministic model also presumes the intersection control provides a high service rate and periodically stops servicing vehicles to accommodate traffic on a conflicting turning movement. This model assumes traffic approaching each intersection as a uniform stream of vehicles arriving at a constant rate [Dion et al., 2004]. The next assumption made in this model is the vehicle decelerates and accelerates instantaneously which allows the conversion of all deceleration and acceleration delays into equivalent stopped delay [Dion et al., 2004]. The deterministic queuing model offers a direct estimation of the total delay incurred by vehicles attempting to traverse an intersection. The last assumption is all the vehicles in the queue are at the intersection stop bar allowing an unbiased delay estimation process over an entire queue formation and dissipation process [Dion et al., 2004]. Equation 2.2 provides the average uniform delay incurred at every signal cycle by vehicles attempting to cross the intersection [Dion et al., 2004]: 26 2 2* er sd C s v ? ?= ? ? ?? ? (2.2) where: d = average delay per vehicle (s/veh), C = traffic signal cycle length (s), re = effective red interval duration (s), s = saturation flow rate (veh/h), and v = vehicle arrival flow rate (veh/h). The methodology computing delay offers greater accuracy if the cycle length for each intersection along the urban street is known. Some of the input parameters can be default values which represent reasonable values for operating parameters. In the advent of unavailability of intersection cycle length, the HCM suggests default values for cycle length based on area type. The default cycle length is 70 seconds for the CBD and 100 seconds for other types of areas [HCM, 2000]. Also according to HCM, the default value of adjusted saturation flow rate per lane for through lanes is 1700 veh/h/ln for the CBD and 1800 veh/h/ln for other types of areas [HCM, 2000]. The traffic on the intersection approach (link) for the calculated hour is the vehicle arrival flow rate. The method used by Hedayat and Iravani (1999), to estimate the red signal time of an intersection approach, is useful for calculating delays when the input data for an intersection is not readily available [Dion et al., 2004]. They have used the following 27 method to estimate the red time of an intersection approach. A j node is an intersection which is equipped with a traffic signal. sj is the set of network links ending at node j. wij is the weight of a link entering node j based upon their functional classification. The following are weights for wij [Dion et al., 2004]: ? wij = 2, if the link is local or collector street, ? wij = 3, if the link is minor arterial, and ? wij = 4, if the link is major arterial. The red time of an intersection is calculated from Equation 2.3 [Dion et al., 2004] as: | | *1.2* * 1 2* j j ij ij i s s wr C w ? ? ? ? ?= ? ? ? ? ?? ?? (2.3) where: r = red interval duration (s), C = traffic signal cycle length (s), sj = the set of network links ending at node j ,and wij = weight of the a link entering node j. The functional form of the Equation 2.3 has the following characteristics [Dion et al., 2004]: 1. Each link of any intersection takes into account the effect of all other links belonging to the same intersection. 2. Intersections with more legs and with higher functional classification will cause more delay time. 28 3. At any given intersection, legs with higher functional classification will take less red cycle time compared to other legs of the same intersection. Geometric, traffic, and signal data of the intersection under study are needed for calculating the delay at a signalized intersection according to HCM [TRB, 2000]. The geometric data includes the number and width of the lanes, parking conditions on the intersection approach, approach grade, and existence of exclusive left turn lanes. Traffic data includes the demand volume by each movement in the intersection, base saturation flow rate, peak-hour factor, percentage of heavy vehicles, approach speed and approach pedestrian flow rate. Lastly, the signal data includes the cycle length, green time, and yellow-plus-all red changes, and clearance interval. 2.5 DELAY FUNCTION FOR UNSIGNALIZED INTERSECTIONS For their traffic simulation analyses Kakooza et. al., (2004) assumed that all turns corresponding to each link entering an intersection have the same delay [Kakooza et al., 2004]. Secondly, they have also assumed that the delay associated with an intersection for each entering link depends on the physical characteristics and control policy of that intersection as well as the volume of traffic on that link [Kakooza et al., 2004]. By using the second assumption Kakooza et. al., did not consider the traffic data from the other links of the same intersection in their delay calculation. This assumption helps to increase the accuracy of a roadway network model when there is limited software capability and data insufficiency. 29 Equation 2.4 provides the delay at an unsignalized roadway intersection [Kakooza et al., 2004]: ( )2 * 10.05 2n kd ?? ?= ?? ? ? ? (2.4) where: d2 = delay at an unsignalized roadway intersection (min), n = number of links ending at the intersection, k = number of links exiting from the intersection. n*(k -1) = the number of possible turns at the intersection. The characteristics of the functional form of Equation 2.4 include (1) intersections with more legs will cause more delay time, and (2) the effect of all other links belonging to the same intersection is considered by each link of any intersection. Geometric data, hourly turning movement volumes, heavy vehicle percentages, pedestrian data, and upstream signal data of the intersection under study are needed for calculating the delay at a stop controlled intersection according to HCM [TRB, 2000]. Geometric factors include number and use of lanes, existence of a two-way left turn lane (TWLTL) or raised or striped median storage (or both), and an approach grade. The presence of traffic signals upstream from the unsignalized intersection under study will produce nonrandom flows and affect the capacity of the unsignalized intersection. 30 2.6 GEOGRAPHIC INFORMATION SYSTEMS (GIS) Geographic Information Systems (GIS) is a multi-faceted discipline built upon many tools and concepts. Based on the U.S. Geological Survey (USGS), ?GIS is a computer system capable of capturing, storing, analyzing, and displaying geographically referenced information; that is, data identified according to location? [USGS, 2006b]. According to the Environmental Systems Research Institute, Inc. (ESRI) ?GIS is a collection of computer hardware, software, and geographic data for capturing, managing, analyzing, and displaying all forms of geographically referenced information? [ESRI, 2006e]. GIS is a rapidly growing technological field that incorporates graphical features with spatially connected data in tabular form to assess real-world problems. The GIS field began with the discovery that a map could be programmed using simple computer code. The map can be modified whenever necessary as it is stored in a computer. This was a massive change from the earlier eras of cartography when maps had to be created painstakingly by hand. In this old cartography method, minor changes required the creation of a new map. The earliest version of GIS was known as computer cartography. The key word for this technology is Geography (i.e., data that referenced to real locations on the earth). GIS can handle and process geographically referenced data. This is achieved by referencing a location by means of longitude, latitude, and elevation. In lay terms, GIS can be thought of as a high-tech map. The spatial data (i.e. a map) is coupled with information in a tabular format known as attribute data. Attribute data is generally defined as 31 additional information linked to spatial data. It is the partnership of these two data types that enables GIS to be such an effective problem solving tool. The capabilities of GIS have evolved greatly from those simple beginnings of computer cartography. GIS produces maps quickly and efficiently, and it stores data in an easily accessible digital format enabling complex analysis and modeling not previously possible. GIS is able to relate different information in a spatial context in order to reach a conclusion about these relationships. Computer cartography, or mapping, is the simplest operation of GIS. The real power of GIS is its ability to use spatial and statistical methods to analyze attribute and geographic information. The end result of a GIS analysis can be derivative information, interpolated information, or prioritized information. GIS allows the visualization of relationships, patterns, or trends intuitively, which may not be possible to see with traditional charts, graphs, and spreadsheets. For many years, GIS has been considered too difficult, expensive, and proprietary. The advent of the graphical user interface (GUI), powerful and affordable hardware and software, and public digital data has broadened the range of GIS applications and brought GIS to the public at large. GIS involves simple line work to represent land features. Line work evolved into the concept of overlaying different mapped features on top of each other to determine patterns and causes of spatial phenomena. Each individual map is called a layer in a GIS system. Figure 2.2 shows the real world at the bottom layer, with five additional layers over it, representing many real world circumstances. The points on the first, or top, layer represent the building locations, and the lines on the second layer represent 32 the roads of the real world roadway network. The third layer shows details about contamination while the fourth layer depicts the shape of the land, and the fifth illustrates different types of land cover. Figure 2.2 GIS Map Layer System Source : [ESRI, 2006f] From the above discussion, one can understand that a GIS map is made up of layers, or a collection of geographic objects that are similar. Layers may contain features or surfaces (i.e., vector data or raster data). Geographic objects vary in shape and size and can be represented as a polygons (e.g. a football field, countries), lines (e.g. roadways, rivers) and points (e.g. home, cities, traffic signals). Geographic objects having discrete features can be represented as vector data. Surfaces have continuous features and continuous numeric values (i.e. measurable values for a particular location, such as elevation, slope, and temperature) rather than shapes. Raster is the most common type 33 of surface data. A raster is a matrix of identically sized square cells. Each grid cell represented in raster data corresponds to the characteristics of a spatial location. The map overlay operation of GIS combines spatial and attribute data from different maps, and/or variables in the map, into a single composite map. Each map feature on the composite map will represent a select set of data characteristics by location. This allows the composite map to be processed further in order to extract new information from the same spatial location for GIS modeling processes. A model is a simplified representation of a real life phenomenon or system. Effective GIS modeling requires the development of analytical models using a GIS system containing spatial data, associated attribute data, and a good data set. Inaccurate data may result in inaccurate models and maps, which will skew the results of the analysis and eventually may result in poor decision making [ESRI, 2006c]. Geographic data comes in three basic forms: 1. Map Data: the location and shape of geographic features. Points, lines, and areas (i.e. polygons) are three basic shapes that represent real-world features. [ESRI, 2006a] 2. Attribute Data: the descriptive data linked to GIS map features is known as attribute (tabular) data [ESRI, 2006a]. Attribute data is collected and compiled for specific areas by obtaining the databases from various organizations. 3. Image Data: data ranging from satellite images and aerial photographs, to scanned maps are considered image data [ESRI, 2006a]. The image data should be converted 34 from a printed to digital format. These images should be geographically referenced to use as GIS data. A final type of data attached to geographic data is metadata. Metadata can be called data about the data [ESRI, 2006d]. Metadata is additional information (besides spatial and attribute data) that is required to make data useful [ESRI, 2006d]. Metadata may include: 1. An inventory of existing data, 2. Definitions of the names and data items, 3. A keyword list of names and definitions, 4. An index of the inventory and the keyword list for access, 5. A record of the steps performed on the data including how it was collected, 6. Documentation of the data structures and data models used, and 7. A recording of the steps used on the data for analysis. Spatial metadata is important because it not only describes what the data is, but it can also reduce the size of spatial data sets. [ESRI, 2006d]. Like any other technology, GIS can be divided into five major components. These components include: 1. Hardware: GIS hardware includes the computer and the operating system to run the GIS software. This includes a variety of computers (e.g. IBMTM, MacintoshTM, UNIXTM) and operating systems (e.g. Windows XPTM ). Digitizers are the most common device for extracting spatial information from 35 maps and photographs into a GIS map. Other hardware includes large/small format printers, digital cameras, and scanners. 2. GIS Software: GIS software includes the program and the user interface for driving the hardware. Widespread user interfaces in GIS include menus, graphical icons, and commands. Table 2.2 presents a list of GIS software producers and their main products. The main product from ESRITM Inc. is ArcGISTM. ArcGISTM 9.1 is the current release of this software at the time of for use in this research. Table 2.2 GIS Software Producers and Software Products. GIS Software Producers Website Main GIS Products Environmental Systems Research Institute, Inc (ESRI) http://www.esri.com/ ArcGISTM Autodesk Inc. http://usa.autodesk.com/adsk/servlet/home AutoCAD MapTM Baylor University in Waco, Texas http://grass.baylor.edu/ GRASS TM Clark labs http://www.clarklabs.org/ IDRISITM MapInfo Corporation http://www.mapinfo.com/ MapInfoTM Intergraph Corporation http://www.intergraph.com/ MGE TM, GeoMediaTM Caliper Corporation http://www.caliper.com/ TransCAD TM, MaptitudeTM 3. Infrastructure: The infrastructure refers to all required physical, organizational, administrative, and cultural environments for GIS operations. The infrastructure includes general organizational patterns, data standards, data clearinghouses, and people with requisite skills. People are needed to design applications, utilize results, and establish criteria for assessments. 4. Data: There are many sources from which GIS data is available. GIS data can be generated in-house using paper maps and information about the contents in 36 the map, digitizers, and scanners. GIS data can be collected using Global Positioning Systems (GPS) and Remote Sensing. GIS data can also be obtained from data providers and the Internet. 5. Methods: To maintain the compatibility between many GIS software programs and to represent the real world in a data model, there are standard procedures for GIS use. There are many accepted procedures for data analysis. 2.6.1 Map Projections and Coordinate Systems The location of the earth features or spatial features are based on a geographic grid. The geographic grid is expressed in longitude and latitude. It is also referred to as geodetic coordinate systems. Figure 2.3 shows the earth divided into latitude and longitude features. Figure 2.3 Earth Divided into Latitude and Longitude. Source : [Raben Systems, Inc, 2006] 37 The features projected on the map are based upon a coordinate system. The map projection translates the location of a geographic grid to a coordinate system or geodetic coordinate system. This system allows positions on the earth?s surface to be described in terms of latitude, longitude, and elevation. Latitude is the angle measured at the center of the earth northwards (or southwards) between the equator and the position of a point on the earth?s surface. The earth?s equator is considered as zero degrees latitude and it is the line from which all other lines of latitude are measured. Latitude is measured between zero degrees to 90 degrees north or south of equator. Longitude is the angle measured at the center of the earth eastwards (or westwards) from the Greenwich meridian or Prime meridian to the position of a point on the earth?s surface. The meridian line is an imaginary line which runs from the North Pole to the South Pole. By international convention, the Greenwich meridian runs through Greenwich of England. The Greenwich meridian is considered as zero degrees longitude and it is the line from which all other longitude lines are measured. Longitude is measured between zero degrees to 180 degrees east or west of Prime meridian. An ellipsoid surface is a mathematical shape model which averages out the shape of the earth. It is also called a spheroid. The height of a feature on the earth?s surface is measured above or below the surface of a mathematical model. Figure 2.4 shows the measurement of latitude, longitude, and elevation of a feature located on the earth?s surface. 38 Figure 2.4 Measurement of an Earth Feature. The shape of the earth is very complex. It is certainly not a perfectly round, spherical planet. It actually bulges outwards along the equator. The actual surface of the earth has lot of undulations caused by hills and valleys. A geoid is an equipotential (equal gravity) or gravitational surface and is equivalent to mean sea level. Every point on the geoid surface is perpendicular to the local plumb line. A geoid can be defined physically and has a complex surface. A geoid can be described by an infinite number of parameters and sensed by instruments. On the other hand, an ellipsoid surface is a mathematical shape model which averages out the shape of the earth. Ellipsoids have a mathematical definition and simple geometrical surface. The ellipsoid can not be sensed on earth surface by instruments. The ellipsoid can be described by two Ellipsoidal Normal Equator Semi Minor Axis Pole Ellipsoid Surface Semi Major Axis Point P Geodetic Longitude at Point P Tangent to Ellipse at Point P Geodetic Height of Point P Greenwich Meridian Geodetic Latitude at Point P 39 parameters, semi-minor axis (polar radius) and semi-major axis (equatorial radius). Figure 2.5 shows the representation of a geoid, ellipsoid, and mean sea level. Figure 2.5 Model of the Earth. Source : [ESRI, 2006b] The datum represents a reference model of the earth and is the basis for a coordinate system. It is derived from an ellipsoid. Many datums have been developed to describe the ellipsoid. For example the North American Datum 1983 (NAD83), Bermuda 1957, South American Datum, and World Geodetic System 1984 (WGS 84). The WGS84 datum uses a GRS-80 ellipsoid and is the latest and overall best fitting ellipsoid of earth. WGS 84 is almost identical to NAD83. It is a satellite determined ellipsoid. The height measured from a GPS instrument is based upon WGS datum. Figure 2.6 shows the measurement for height of an earth?s feature with respect to an ellipsoid surface. 40 Figure 2.6 Earth Surfaces Based on GIS. A map is a flat surface usually representing a curved surface. Transformation of the locations on the earth onto the flat surface of the map is called map projection. It is easier to work with two dimensional coordinates rather than three dimensional spherical coordinates. The map projection enables map users to work with two dimensional coordinates. All map projection processes distort at least one property (area, distance, or direction) of the earth?s features. If a relatively small area is being mapped, then this distortion is negligible. A fundamental principle of GIS is that the user has to use the same projection for all the layers in the map to conduct analysis among them. Essentially a map projection is developed by taking a flat piece of paper, and forming it into the shape of a cone or a cylinder around the earth. The paper will touch at one line of earth, and all the other parts of the earth are projected onto the paper. The line of tangency where the paper touches the earth is either called a standard parallel (latitude) Ellipsoid Surface Topographic Surface Geoid Surface h 41 or a standard meridian (longitude). The map projection is categorized as a conic projection, a cylindrical projection, or an azimuthal projection depending on wheather it was constructed using a cone, cylinder, or plane, respectively. The azimuthal projection has a point of tangency instead of line of tangency. Figure 2.7 shows the three most important projections cylindrical, conic, or azimuthal. A Transverse Mercator projection uses a standard meridian as line of tangency. This projection places the cylinder in a horizontal position around the earth. A Lambert Conformal Conic Projection is based on conic projection and has two standard parallels. The lambert conformal projection is suited for the mid-latitude areas of the earth. Figure 2.7 Geometric Representations of Projections. The coordinate system is based on the map projection principle. Coordinate systems are designed for detailed calculations and positioning on the earth surface. The absolute position and the relative position accuracies of a feature on earth are very important. A coordinate system is divided into different zones to maintain the accuracy of the measured quantity. Each of these zones is based on separate map projections. Commonly used coordinate systems are the Universal Transverse Mercator (UTM) grid Conic Cylindrical Azimuthal Transverse Mercator 42 system and the State Plane Coordinate (SPC) system. Alabama maps are usually displayed in the SPC system. It is primarily used due to its accuracy in terms of linear measurement. A state may have two or more SPC zones to maintain an accuracy of one part in 10,000 or less. Figure 2.8 shows the zones for all states in SPC system. This coordinate system is very practical while working with small areas, such as cites or towns. Where as this system may not be adequate when dealing with larger areas such as states or countries. The east side of Alabama is projected in SPC system called ?NAD 1983 State Plane Alabama Feet?. The designation ?feet? is added at the end of this coordinate systems name to show the unit of measurement for the projected map. Figure 2.8 Zones of State Plane Coordinate (SPC) System in the USA. Source : [Dean J. D., 2006] 43 2.6.2 Digital Orthophoto Quad (DOQ) According to U.S. Geological Survey a Digital Orthophoto Quadrangle (DOQ) is a computer-generated image of an aerial photograph in which image displacement caused by terrain relief and camera tilts has been removed. A DOQ combines the image characteristics of a photograph with the geometric qualities of a map [USGS, 2006a]. A conventional perspective aerial photograph contains image distortions caused by the topography and camera tilt. In an aerial photograph distances can not be measured because they lack uniform scale. An orthophoto is a uniform-scale image which allows distances to be measured directly from the image similar to scaled cartography maps. A DOQ is geo-referenced and can serve as a base map or background for updating existing GIS maps. The DOQs are square in size and contain the ground size of 3.75 minutes image. 3.75 minutes is the equivalent ground length of 3.75 minute longitude and latitude. Each pixel of the 3.75 minute digital map will measure and represent 1m x 1m on the ground. Georeferenced aerial photos are also orthophotos, but they have ground length less than 3.75 minute longitude and latitude. 2.7 ARCGISTM SOFTWARE Environmental Systems Research Institute, Inc. (ESRI), has developed the ArcGISTM software which contains three main subcategories: ArcMapTM, ArcCatalogTM, and ArcToolboxTM. ArcMapTM is the central application in ArcGISTM Desktop for all map-based tasks including cartography, map analysis, and editing. ArcMAPTM has the ability to create maps from layers of spatial data. The ArcCatalogTM application organizes and manages all GIS information such as maps, 44 models, data sets, and metadata. It can be used for browsing the data on a hard disk, network, or the Internet. It can also be used to search spatial data, preview the data, and add the data to a map in ArcMAPTM. ArcToolboxTM is embedded in ArcCatalogTM and ArcMapTM. ArcToolboxTM contains a comprehensive collection of geoprocessing functions which are tools used for changing the projection of a map, and exporting and importing data from other types of GIS software. ArcGISTM uses a shape file data structure which is composed of at least four types of files. The file types include are shape files, shape index files, dBASE files and a spatial projection information file and contain the file extensions *.shp, *.shx, *.dbf, and *.prj, respectively. The shape file stores the feature geometry. The shape index file stores addresses to the feature geometry for faster access and operation of the shape file. The dBASE file stores the attribute information of features, while the projection file stores the projection information of the map. The visual basic (VB) programming language was developed by MicrosoftTM based on an object-oriented language and is intended for application development. GIS attributes and map data can be processed using VB script. It provides GIS operators the flexibility to run or perform various decision-making process algorithms on attribute data. This feature is very helpful for GIS modeling. ArcGISTM will have several extensions for additional applicability. The ArcGIS Network AnalystTM provides network or grid based spatial analysis. This extension can be used to design analytical models to find the best route based on specified road 45 network attributes such as time and distance. It can generate directions of the route with turn-by-turn maps of the network. This extension can be used to find the areas that fall within certain travel times, distances, or costs from a facility. Network analyst is a widely used application of GIS and is used by emergency services, utility companies, regional transportation authorities, railway companies, and other city services. This type of analysis can be used to model hydrologic flow, traffic flow, the shortest and fastest routes, and delivery routes. Roads, pipelines, sewer lines, and rivers networks benefit from the capabilities of GIS network analysis. 2.8 GLOBAL POSITIONING SYSTEMS The Global Positioning System (GPS) is a satellite based navigational system used to locate a position anywhere on the surface of the earth with a help of GPS instruments. This technology is available to every one, everywhere on earth, day and night without any cost for navigational data. This technology can be used in any application that requires location or measurement. GPS uses a constellation of 24 satellites revolving the earth in an orbit of 11,000 nautical miles [MSUB, 2006]. There are 4 satellites in each of six orbital planes. Each orbital plane is inclined 55 degrees relative to the equator. Figure 2.9 shows the constellation of GPS satellites. These satellites are operated by the U.S. Department of Defense (DoD), and is known as NAVSTAR (NAVigation Satellite Timing and Ranging). This constellation of satellites will allow GPS receivers to acquire signals from at least four satellites from any point on the earth, at any time. The high altitude of 11,000 nautical miles ensures the satellite orbits are stable, precise, and predictable [Cooksey, 2006]. GPS satellites provide an 46 accurate timing system by using highly accurate atomic clocks mounted on them. GPS satellites continuously broadcast satellite position and timing data by means of radio signals. The GPS satellite radio signals cannot penetrate water, soil, or other obstacles; and the signal follows a straight line. Figure 2.9 Constellation of GPS Satellites Around the Earth. Source : [Garmin Ltd. 2006] 2.8.1 How GPS Works A GPS instrument or receiver on earth calculates its position by measuring the distance between the GPS receiver and the GPS satellites. The position of each satellite is known as it orbits the earth in a prescribed path. The satellites transmit their position and timing messages as part of the information they transmitted via radio waves. The GPS receiver on the ground is at the unknown point. GPS receivers receive radio signals from the satellites. All GPS receivers are synchronized with the satellites so they generate the same digital code at simultaneously [Cooksey, 2006]. When the GPS receiver receives a code from a satellite, the receiver calculates the difference in time of signal reception and signal emission. The difference in time is multiplied by the speed 47 of light, which is approximately the travel speed of the signal emitted by the satellites, to obtain the distance between the GPS receiver and satellite. Figure 2.10 illustrates the time difference between the radio waves received and generated by a GPS receiver. Figure 2.10 Time Difference Between Radio Waves. Source : [Cooksey, 2006] A GPS receiver must receive signals from at least three satellites to calculate its latitude and longitude position. When a GPS receiver can obtain signals from four or more satellites, the GPS receiver has the ability to determine its latitude, longitude, and altitude position while also minimizing the error in calculating its position. For a moving GPS receiver the speed and path of the receiver can be calculated by knowing various positions along its path and the time at which it occupied the position. The positional measurement of an object with a GPS receiver will usually have a few meters of error. While measuring the path and speed of a moving object, the GPS instrument GPS Receiver Code received from the satellite Code sent by the satellite and generated by the receiver GPS Satellite Time Difference 48 measures the relative position of the object. Therefore the GPS instrument will exhibit more error while measuring a fixed position of an object in comparison to measuring the speed and path of a moving object. 2.9 SUMMARY It is possible to find the shortest or fastest route with reliable travel time prediction for ready-mix materials for construction projects. For this purpose GIS and various traffic related data of the roadway network is needed. In this research the travel time of each roadway and the delay caused by various types of intersections should be considered during roadway network modeling. For the calculation of the travel time the fluctuation in the traffic should also be considered. GIS software modeling is a better option for identifying the shortest time and shortest distance route due to its excellent visualization capability in comparison to traditional techniques of wayfinding. The scope of this research effort is to find the shortest route and the fastest route along with their associated cost of transport for ready-mix concrete materials to aid decision makers in better route selection. 49 CHAPTER THREE DATA COLLECTION 3 3.1 COLLECTION OF GIS DATA The City of Auburn (COA) Public Works Department was contacted for the collection of GIS data for the Auburn, AL area. Ms. Liesa Simpson, the Public Works Department?s engineering technician, provided the GIS data for the COA and was obtained on February 15th, 2006. It contained three GIS shape files. Table 3.1 describes the three types of shape files collected of the COA. All shape files were in vector data format. Geo referenced aerial photos of the entire Auburn area was also collected. A total of 245 geo referenced aerial photos were collected to cover the entire study area. All the GIS data collected was in NAD 1983 State Plane Alabama East Feet coordinate system. Table 3.1 Description of Shape File GIS Data Name of Shape File Geometric Data Number of Records Description of Field StreetCL Polyline 4803 This file contains street network data for the COA. It is centerline data of the COA streets with associated attribute tables explaining many properties of each street section. AADT2006 Point 460 This file contains location data of Annual Average Daily Traffic (AADT) collection stations with associated attribute tables providing AADT values and the date of traffic data collection. Signals Point 61 This file contains the location of traffic signals in the COA with associated attribute tables providing information on the inventory of traffic signal heads. 50 3.1.1 GIS Streets Data for the City of Auburn Figure 3.1 illustrates the COA roadway GIS data and associated attribute data. Figure 3.1(a) shows the entire COA roadway network utilized for network modeling. Figure 3.1(b) displays a snapshot of the attribute table that contains additional data for the COA roadway network. (a) GIS Map of COA streets (b) Attribute Table of Auburn Area Streets GIS Data. Figure 3.1 City of Auburn GIS Data of the Roadway Network. There are a total of 4,803 records contained in the attribute table. Each record is represented by one row which contains all attribute data associated with one street in the roadway network. The GIS street data provided also contains metadata. After Legend: Street N 51 analalyzing the attribute data and metadata, 9 fields were deemed relevant for transportation network modeling. Table 3.2 provides descriptions for the relevant columns or fields of the attribute data. Table 3.2 Description of Relevant Field Contained in StreetCL Attribute Table Field Name Value Type Description of Field FID Integer Identification number of the row Shape Characters Geometric shape of the feature or street SPEEDLIMIT Integer Speed limit (mph) CLASS Integer Functional classification of the street. 1 = major Arterial, 2 = minor Arterial, 4 = collector street, 5 and 0 = local street STATE_HWY Integer State highway number US_HWY Integer US highway number DIRECTION Character Direction of street. N,S,E,W, for North, South, East, and West respectively NAME Characters Name of the street TYPE Characters Type of the street. RD = road, ST = street, AV = avenue In the attribute data, it has been observed that some local streets did not contain a specified speed limit. The majority of the local roads on which speed limit values were missing are residential streets in the COA. As a result, a minimum speed limit value of 25 mph was assumed for these local/residential streets. 3.1.2 GIS City of Auburn Traffic Data Figure 3.2 shows the COA Annual Average Daily Traffic (AADT) data and associated attribute data. Figure 3.2(a) illustrates the position of the AADT data collection locations overlaid on top of the GIS roadway network for easy and meaningful visualization. There are a total of 460 records of AADT data collection locations for the entire COA roadway network. Figure 3.2(b) shows a sample of the attribute data associated with the AADT data. 52 (a) AADT Data Collection Location on COA Streets. (b) Attribute Table of AADT Data. Figure 3.2 City of Auburn AADT Data. The metadata provided for the AADT shape file properly explained the details about all the field names contained with in the file. Table 3.3 provides the description for the relevant fields of attribute data. Table 3.3 Description of Each Field in AADT2006 attribute table Field Name Value Type Description of Field FID Integer Unique numerical identifier Shape Characters Geometric shape of the feature or street COUNTS Integer AADT value of the street DATE Integer Year designating collection of AADT data Legend: Street AADT value N 53 3.1.3 COA GIS Traffic Signal Data Figure 3.3 shows COA traffic signal data and associated attribute data. In Figure 3.3(a) the points illustrated represent traffic signal locations in the COA which are overlaid on the COA roadway network for easy and meaningful visualization. Figure 3.3(b) shows the attribute data associated with the traffic signal data. (a) COA Traffic Signal Locations (b) COA Traffic Signal Attribute Table Figure 3.3 City of Auburn Traffic Signal Data. The metadata provided for traffic signal shape file did not provide adequate descriptions for all fields contained in the table. The description of many fields were discovered by exploring the attribute data values and names of the field. Table 3.4 provides the Legend: Street Signal N 54 descriptions for the relevant traffic signal attribute data fields used for the GIS Modeling. Table 3.4 Description of Each Field in Signals Attribute Table Field Name Value Type Description of Field FID Integer Number of the row Shape Characters Geometric shape of the feature or street SIGNAL Characters Name of the two streets at the intersection 3.2 INCORPORATION OF CRITICAL COA GIS ROADWAY NETWORK PARAMETERS The GIS data provided of the COA roadway network did not contain all input parameters required for excellent roadway modeling. The AADT shape file did not contain relevant information on the directional distribution of traffic and the percentage of truck volume in the traffic stream. The traffic signal shape file did not contain information on the intersection delay experienced on each link. The DOQ file can be utilized to determine the number of lanes each street within the network contains. AADT values of many streets and intersections containing traffic signals can be incorporated into the street shape file by using individual shape files containing AADT and traffic signal information. 3.2.1 Number of Lanes The COA roadway network GIS file did not contain information describing the number of lanes each street was comprised of. The number of lanes a street contains is a very important parameter in determining the traffic capacity of a street. The capacity of each street along with the current traffic volume each street experiences is essential in 55 determining realistic travel times. With this in mind, the COA roadway network needs to be updated to include information detailing the number of lanes for each street link. To determine the number of lanes, Geo referenced data was utilized to assist in upgrading the roadway network GIS file to include the number of lanes information. Figure 3.4(a) illustrates the COA roadway network GIS data overlaying 110 Geo referenced files. With the digital aerial photograph as the background of the roadway network, the number of lanes for all 4,803 streets could be determined and entered in roadway network attribute table. A new field entitled as ?nooflane? was added to the roadway network attribute table. Figure 3.4(b) shows the attribute data associated the COA GIS roadway network data with the field updating the number of lanes. 3.2.2 Traffic Signal Locations The traffic signal information at intersections was not available in the COA roadway network file. To make the traffic signal information available in the road network GIS file, spatial join operation of ArcGISTM was performed. The information of whether a traffic signal was present at the end of a street or not was obtained from traffic signal GIS file of COA. The answer obtained for the above question is either true or false and the answers are represented as ?1? for true and ?0? for false. Occasionally on some streets in at COA roadway network, traffic signals were present at its both ends; therefore a value ?2? was introduced to represent such street types. 56 (a) City of Auburn GIS Geo Referenced Photos overlaid with Streets Data. (b) Attribute Table of COA Streets GIS Data with Number of Lanes Field Figure 3.4 Number of Lanes Incorporated in the COA Roadway Network GIS Data. Figure 3.5 shows attribute file of GIS street file with presence of traffic signal at the end of the street. A field name ?count_? is assigned to incorporating the values representing the presence of a traffic signal in the attribute table of COA roadway network GIS file. Figure 3.5 Traffic Signal locations Incorporated into the COA Roadway Network. Legend: Street N 57 3.3 TRAFFIC DATA COLLECTION A substantial amount traffic data for the COA roadway network can be collected from the AADT GIS file. Additional traffic information was required for this research to generate the traffic factors and to determine the percentage of growth in traffic experienced by the COA. The collection of this addition traffic information is described in the following sections. 3.3.1 Annual Average Daily Traffic (AADT) Values The AADT values were contained in a separate shape file and need to be incorporated into the street shape file. The analysis on the network to identify the optimized route will be conducted on the street shape file. The spatial join option of ArcGISTM was used to join the AADT and StreetCL shape file. Figure 3.6 shows the attribute file of the updated COA roadway network GIS file with AADT values incorporated. The ?Date_? field provides the information about the year on which AADT was collected on at a particular the location. The ?Integer? field furnishes the AADT information for the street. In the Integer field if numeric value is zero, then it means no traffic data was collected on that street or link. Figure 3.6 AADT values Incorporated Into the COA Roadway Network. 58 3.3.2 Traffic Multiplication Factors Traffic multiplication factors for the COA are required to calculate the hourly traffic distribution on each street in the roadway network. To generate traffic multiplication factors for the COA, the continuous traffic data for a minimum of one year is required. Automatic Traffic Recorder (ATR) count stations are the only means for measuring the traffic a roadway experiences throughout the year. There is one ATR station in the COA located on South College AL 147 near the intersection of Angeland and South College [ALDOT, 2006b]. The number of the station is ATR705. The year round traffic data collected by the ATR705 station was obtained for the years 2002, 2003, and 2004 from the Alabama Department of Transportation. Of the three years of traffic data obtained, the year 2004 data was corrupted. For the years 2002 and 2003, the traffic data for some days of a few months were missing. The hourly traffic percentage traffic for a day is obtained for every hour using the following formula: Traffic at an hour of a dayHourly traffic percentage = *100Total traffic per day (3.1) Traffic for any hour of the day is the traffic that occurred on the roadway in a given hour of time for which the hourly traffic percentage is calculated. The hourly percentage value of each hour is averaged for an entire year. Finally to obtain the hourly percentage traffic factor for each hour, the value of hourly percentages of that hour were averaged over two years. Table 3.5 tabulates the hourly traffic percentage values obtained using traffic data collected by ATR 705 station for each hour on the COA roadway network. 59 Table 3.5 Hourly Variations of Traffic Volumes on the COA Roadway Network Hour Percentage of Total 24- Hr Volume Hour Percentage of Total 24- Hr Volume 12.00-1.00 A.M. 1.735 12.00-1.00 P.M. 6.892 1.00-2.00 A.M. 1.256 1.00-2.00 P.M. 6.606 2.00-3.00 A.M. 1.018 2.00-3.00 P.M. 6.691 3.00-4.00 A.M. 0.645 3.00-4.00 P.M. 7.160 4.00-5.00 A.M. 0.543 4.00-5.00 P.M. 7.467 5.00-6.00 A.M. 0.952 5.00-6.00 P.M. 7.154 6.00-7.00 A.M. 2.317 6.00-7.00 P.M. 6.271 7.00-8.00 A.M. 4.047 7.00-8.00 P.M. 5.345 8.00-9.00 A.M. 3.792 8.00-9.00 P.M. 4.686 9.00-10.00 A.M. 4.427 9.00-10.00 P.M. 4.052 10.00-11.00 A.M. 4.999 10.00-11.00 P.M. 3.346 11.00-12.00 A.M. 6.088 11.00-12.00 P.M. 2.481 Similarly the weekday multiplication factor for all the days of the week was obtained using the following formula. Total traffic per yearMultiplication factor of a weekday = 7 * Total traffic of the weekday per year (3.2) The multiplication factor of a weekday for both years was averaged to calculate the final value. Table 3.6 provides the weekday multiplication factors for the COA. Table 3.6 Daily Variation of Traffic Volume on the COA Roadway Network Day Weekday multiplication factor Sunday 1.196 Monday 1.024 Tuesday 0.988 Wednesday 0.974 Thursday 0.936 Friday 0.870 Saturday 1.060 60 The traffic data for the entire month was not available for the traffic data collected. Therefore it was not possible to determine the seasonal factors by the months for the two years of traffic data available. For this reason the seasonal factors by each month is provided by the Institute of Traffic Engineers (ITE) Traffic Engineering Handbook. These values are furnished in Table 3.7. Table 3.7 Seasonal Variation of Traffic per Month Source: [ITE, 1965] Month Monthly factor Month Monthly factor January 1.215 July 0.913 February 1.191 August 0.882 March 1.100 September 0.884 April 0.992 October 0.931 May 0.949 November 1.026 June 0.918 December 1.148 3.3.3 Percentage of Traffic Growth The AADT traffic data for the COA over the previous year had to be obtained to determine the percentage growth in traffic. The Alabama Department Of Transportation (ALDOT) collects AADT traffic data at various locations on the COA roadway network [ALDOT, 2006a]. The AADT data collected at those locations are available over the Internet. The AADT data for the years 1995 and 2005 were collected from a total of 29 locations in the COA. Compound growth factors per year was assumed to determine the traffic growth occurring in the COA. The percentage growth (also known as the Annual Growth Model (AGR)) for COA roadway network was found to be 2%. 61 The AGR value for the COA was obtained by Equation 3.3: 1 Future AADT AGR 1 Current AADT n? ?= ? ? ?? ? (3.3) where: AGR = annual growth rates (%), n = difference in future and current years of traffic data collection. 3.3.4 GPS Data Logger The GPS data logger used in this research was custom made to collect the GPS data of heavy vehicles such as vehicles used in the agriculture and construction industries. The GPS data logger was designed to withstand the vibration experienced during the operation of large vehicles. The data logger was developed by Matt Darr, Research Associate in the Department of Food, Agricultural, and Biological Engineering at Ohio State University. For this research the GarminTM GPS receivers was used to collect GPS data. GarminTM GPS receivers are accurate within 3 to 5 meters on average [Garmin Ltd., 2006]. The GPS receivers have a sturdy plastic casing. The GPS instruments were mounted on top of concrete haul trucks in a safe position to avoid obstructions so GPS radio signals could be acquired. Figure 3.7 shows pictures of the GPS data logger and GPS receiver. A cigarette lighter plug was used to supply power to the data logger. The GPS data logger processed the GPS data received every second and stored the data into a memory drive in *.txt format. The GPS instrument uses a compact flash drive for storing data. Figure 3.8 shows the compact flash drive mounting area of the GPS data logger. The power switch has an ON or OFF function. When power 62 switch is in on position the instrument will start collecting data. The display panel will show the status of the data collection. Figure 3.7 GPS Data Logger. Figure 3.8 GPS Data Logger Compact Flash Drive Mounting Area. 3.4 GPS DATA COLLECTION FOR CONCRETE TRUCK HAUL ROUTE Mr. Charles Bell of Twin City Concrete Co. located at 214 Twin City Ct, Auburn, AL was contacted for collecting the GPS data of ready-mix concrete trucks. This Power Switch Cigarette Lighter plug Opening for Inserting Memory Drive GPS Data Processing Instrument Germin? GPS Receiver (Clipper) Display Panel Compact Flash Card 63 company is part of Sherman Industries Inc., a diversified producer of concrete ready-mix and concrete products company. The Twin City Concrete Co. supplies ready-mix concrete to locations in the cities Auburn, Opelika, and Waverly. Mr. Bell was very generous to allow the research group to mount the GPS instruments on his ready-mix concrete haul trucks. He also provided additional information on the time and place of dispatch for each trip in which the GPS instrument was mounted to the ready-mix concrete trucks. At Twin City Concrete Co. the ready-mix concrete is transported by front discharge OSHKOSH? s-series concrete trucks. The model of the truck is an AS2446 and the manufacturer date or make of the truck is 2005. It has CaterpillarTM manufactured C-13 engine and McNeilus made concrete mixer drum. The empty weight of the truck is 29,000 pounds. The trucks carry 9 cubic yards of concrete in a fully loaded condition weighing 68,000 pounds. Figure 3.9 is an illustration of the ready-mix concrete truck used in the research. GPS Receiver Mountable places 64 Figure 3.9 Ready-mix Concrete Truck. Two GPS instruments were available during the research effort and were mounted on two separate ready-mix trucks. The scaffolding provided around the concrete receiving funnel of the ready-mix truck was the highest point on the truck providing a suitable location to mount the GPS receivers. The GPS receivers were mounted on the right side of the trucks. Figure 3.10(a) and Figure 3.10(b) shows the GPS receivers mounted on the ready-mix concrete trucks. Driver Side Door GPS Receiver Mountable Places 65 (a) On Right Side of Truck No. 1 (b) Right Side of Truck No.2 Figure 3.10 GPS Receivers Mounted on Ready-mix Concrete Trucks. Once the GPS receivers were mounted in proper locations on the trucks the receivers were connected to the power source. Inside truck cabin the GPS instruments were placed on a firm surface. Figure 3.11 shows the GPS instrument placed safely inside the driver?s cabin with all its connections secured. Figure 3.11 GPS Instrument Mounted Inside the Truck. Cigarette Lighter Plug in its Socket places GPS Data Logger Cord From Mounted GPS Receiver GPS Receiver GPS Receiver 66 3.4.1 GPS Data Collection GPS data was collected for two trucks hauling material between 25th of May 2006 and 16th of June 2006. The GPS data was collected over 17 working days. The drivers of the haul trucks had been directed to switch ON the GPS instrument in the morning when their working day begins and switch it OFF when their work for the day ends. The GPS data of the ready-mix trucks were collected on the GPS data logging instrument throughout the day. The GPS data collected for the truck hauls were projected on the COA GIS roadway network map. The data on haul routes followed to deliver concrete to Waverly and Opelika which were outside the COA limits were disregarded. Only the routes in which the entire truck haul was in the COA limits were considered for GIS modeling since researchers only had traffic information on COA streets. Of the entire valid GPS data collected, a total of 36 routes were found to be adequate for GIS modeling. Of those 36 routes, only the portion of the route that represents a haul or a return haul on the GIS roadway network were considered. Therefore the portion of the haul that went of the roadway network and onto the construction site was not considered. The haul and return routes were determined using the time of the data collection information provided in the GPS data. Figure 3.12 shows the GIS roadway network and a portion of the network near the batch plant location. The location of the batch plant is fixed for every haul route. The time that the GPS data collected on the short link (i.e. 67 batch plant access road) in the network connecting the batch plant to roadway network is studied for every route. Figure 3.12 GIS Road Network at Batch Plant Location. Since location A is closer to the batch plant and location B is farther from the batch plant it was necessary to examine the GPS timings data to determine whether the concrete truck was in an haul or return condition. If the time of GPS data point at location A is Batch Plant Location Location B Batch Plant Access Road * Location A N N 68 earlier to the time collected for location B, the route is categorized as an haul condition. Similarly if the time of GPS data point at location A is later than the time collected at location B, the route is categorized as return condition. 3.4.2 Collected GPS Data The GPS data was stored in a *.txt format file on the flash card mounted inside the data logger. The *.txt file was in a comma separated file format. Figure 3.13 illustrates the raw GPS data obtained from the data logger. Figure 3.13 GPS Data Stored on the GPS Data Logger Instrument Flash Card. The raw GPS data contains several rows and each row represents the GPS data collected every second during the GPS data collection period. The raw GPS data also contains several columns and each column represents a particular type of data. The relevant data types stored in columns are explained in Table 3.8. Table 3.8 Column Descriptions of Raw Data Data Type Data Description DATE Date on which the GPS data was collected in mm-dd-yy format TIME Time at which the GPS data was collected hh:mm:ss format LAT DEG Degree value of the latitude position of the GPS receiver at that time in integer format. LAT DEG MIN Minute value of the latitude position of the GPS receiver at that time in decimals. LON DEG Degree value of the longitude position of the GPS receiver at that time in integer format. LON DEG MIN Minute value of the Latitude Position of the GPS receiver at that time in decimals. SPEED(KM/HR) Speed at which the vehicle carrying the GPS instrument is traveling at that time as a integer. 69 The GPS data is transferred from the compact flash card to a computer using a compact flash card reader. The GPS data in *.txt format cannot be directly used in the ArcGISTM software for modeling purposes. To use the GPS data collected in the ArcGISTM software it has to be converted into a shape file format. During the GPS data collection of the trucks for a route, GPS truck location data is stored for each second. In each GPS data file the time, latitude, and longitude of the truck was collected for every point recorded. Figure 3.14 shows the projection of the collected GPS data overlaying the COA roadway network. In the Figure 3.14 the GPS points are clearly visible on a small scale map of the COA GIS street network. For an haul or return route, the time difference between the GPS data collected at the beginning and at the end of the route is the actual travel time for the truck. As time data is collected in hours, minutes, seconds (i.e., hh:mm:ss) format, the actual travel time is calculated with seconds accuracy. Due to the frequency of GPS data collection, when the GPS data points are projected on a larger scale on the GIS map, the points resemble a line. The line represents the haul route actual taken for delivering the ready-mix concrete. 70 Figure 3.14 Representative Haul Route GPS points on GIS Road Network. 3.5 CONCLUSION The GIS data for the COA relevant for the roadway network modeling was collected. GIS data including: information about the roadway network, AADT values of Location of Batch Plant Ending Point of Haul * Starting Point of Haul GPS Point N N 71 the network and traffic signal details were collected. The traffic data collected was adequate to find the percentage growth rate of traffic and to also obtain traffic multiplication factors for the COA. The GPS data collected for the described trucking operation consisted of 36 concrete truck haul routes. The GPS data furnished the vital information for calibrating the GIS roadway network models (i.e., travel time of trucks for haul/return routes and to optimize route selection for the delivery of the material). 72 CHAPTER FOUR GIS MODEL DEVELOPMENT, VALIDATION, AND OPTIMIZATION 4 4.1 INTRODUCTION In this chapter Geographic Information System (GIS) models are being developed to find the shortest distance and shortest time routes for the delivery of material by trucks. When the trucks travel along the shortest distance or time routes, savings of time and money for the entire operation should be experienced. The GIS models will be validated and calibrated by comparing the actual truck travel data with model generated data to determine if travel times are accurate before running the shortest distance and shortest time models. Once the reliability of the GIS models developed is established, savings in terms of money and time can be determined. The required travel time for a vehicle traveling from one location to another on a roadway network is the cumulative travel time required to travel each link and traverse each intersection on which the vehicle travels. The attribute table of the GIS file used for model development contains 4,805 rows. Each row provides attribute information for one street link. By using the attribute values contained in the GIS road network attribute file, users can only calculate link travel times and are not able to calculate intersection 73 delay. Therefore the total delay experienced at an intersection by the vehicle traveling through it is required, and can be obtained by incorporating the intersection delay to all adjoining links. Intersection delay was considered during the model development process and the delay experienced by traveling through the intersections was determined. Intersection delay was added to all links adjoining at a particular intersection. Now by combining the travel time required to traverse a link with relevant intersection delay information, the total actual travel time required to travel on the roadway network can be determined. Using and deriving information from the available GIS data set, it is not possible to develop the desired time-dependent GIS models for the COA Roadway network. Therefore some assumptions were made by the researchers to develop the time- dependent GIS models. Table 4.1 lists the assumptions made in this research work and each assumption is explained in detail in sections were each one them is applied. Table 4.1 Assumptions for Development of Time-Dependent GIS Models Assumption 1 Base Free flow speed on the link is same as the speed limit value of the roadway link. Assumption 2 Wherever the speed limit value is not provided for the local streets by the GIS data, a minimum speed limit value of 25 mph is assumed. Assumption 3 According to HCM guidelines, 100 seconds for traffic signal cycle length and 1800 veh/h for saturation flow rate at traffic signals is applied. Assumption 4 A direction distribution of 50% is assumed for the traffic in the entire COA roadway network. Assumption 5 ITE monthly traffic variation factors are applied to traffic in COA roadway network. Assumption 6 Vehicles traveling on a major and minor arterial will not experience any delay at un-signalized intersections joining the major and minor arterials to local and collector roads. 74 4.2 BASE GIS MODEL DEVELOPMENT METHODOLOGY The GIS model developed in this section represents the base model. The base GIS model is will incorporate all the basic model parameters developed prior to the inclusion of the traffic multiplication factors. All of the basic model parameters will be same for every model and are not time-dependent. However the traffic multiplication factors will vary hourly for any given day, therefore separate time-dependent GIS models need to be developed to incorporate these model specific factors for each haul route. 4.2.1 Signalized Intersection Delay For effectively developing a roadway network model the time required to travel the roadway and delay caused by signalized and unsignalized intersections is important. Equation 2 discussed in the literature review can provide the average delay experienced by vehicles at signalized intersections is illustrated below as [Dion et al. 2004]: 2 2* er sd C s v ? ?= ? ? ?? ? (4.1) where: d = average delay per vehicle (s/veh), C = traffic signal cycle length (s), re = effective red interval duration (s), s = saturation flow rate (veh/h), and v = vehicle arrival flow rate (veh/h). 75 From the above equation it is evident that in order to calculate delay at signals the information on the traffic signal?s cycle length, red time, saturation flow rate, and vehicle arrival flow rate is required. Of these parameters, vehicle arrival flow rate can be derived from using Annual Average Daily Traffic (AADT) information on the City of Auburn (COA) GIS roadway network. Based on the Highway Capacity Manual (HCM) guidelines, saturation flow rate and traffic signal cycle length can be assumed constant values. The red signal time can be derived from available data at contained in the COA GIS roadway network file and assuming traffic signal cycle times. The methodology for deriving the red signal time and vehicular flow rate is explained later in this chapter. The GIS network does not provide metadata or attribute information to differentiate between the Central Business District (CBD) and other areas in the COA roadway network. Therefore in this research the entire COA roadway network was considered as an area other than CBD. According to HCM, guidelines for traffic signal cycle length and saturation flow rate can be assumed as 100 seconds and 1800 veh/h respectively for areas other than the CBD. Two new fields were added to the attribute table labeled as cycle length (i.e.?cycle_leng?) and signal saturation (i.e. ?Signal_sat?). A value of 100 was placed in the ?Cycle_leng? field using the VB script presented in Table 4.2. In a similar fashion, a value of 1800 is placed in the ?Signal_sat? field by replacing the value 100 with 1800 in the fourth line of the VB script presented in Table 4.2. In Figure 4.1 cycle length values of ?100? or ?0? are inserted depending upon whether a signal is present or not ( i.e. ?count_= 1 indicates the presence of the signal) 76 Table 4.2 VB Script Used to Update Traffic Signal Cycle Length Code Dim cycle as Integer Dim Count_ as Integer If [Count_] > = 1 Then cycle = 100 Else cycle = [Count_] End If Value in Result Field Cycle Figure 4.1 Attribute Data Till Projection of AADT Once the traffic signal cycle length is known, link weights can be assigned to all the links (streets) entering the intersection, so the red signal time of the intersection can be calculated. Equation 2.3 [Dion et al., 2004] discussed in literature review chapter is utilized to calculate each intersection?s red signal time and is provided below as: | | *1.2* * 1 2* j j ij ij i s s wr C w ? ? ? ? ?= ? ? ? ? ?? ?? (4.2) where: r = red interval duration (s), C = traffic signal cycle length (s), sj = the set of network links ending at node j ,and wij = weight of the a link entering node j and wij = 2, if the link is local or collector street, 77 wij = 3, if the link is minor arterial, and wij = 4, if the link is major arterial. The ?class? field of the attribute table provides a description of each street functional classification where: 1 was assigned for major arterials, 2 for minor arterials, 4 for collector streets, and 5 or 0 for local streets. A new data field was created in the attribute table with the name as ?weight1?. The VB script in Table 4.3 was created to change the link value based upon the type of road into the required weight system format required by Equation 4.2 for each link. Table 4.3 VB Script of for Weight of The Link Code Dim class as Integer Dim weight1 as Integer If ([class] < = 0 or [class] > = 5 ) Then weight1 = 0 ElseIf [class] = 4 Then weight1 = 2 ElseIf [class] = 2 Then weight1 = 3 ElseIf [class] = 1 Then weight1 = 4 EndIf Value in Result Field weight1 A new field labeled as ?Red_min? is added to the attribute table providing the red interval for all signalized intersection within the network. Now the weight of the every link entering a signalized intersection and cycle length for all traffic signals is available for all signalized intersections. However the information describing the number of nodes ending at an intersection and the classification of roads terminating at an intersection can not be retrieved merely by refereeing the attribute table. These two data sets were only found by inspecting each signalized intersection manually using the GIS 78 map and retrieving relevant information from the attribute table simultaneously. The red signal time value for each link entering all the 61 signalized intersections in the COA roadway network were determined and manually entered into the attribute table. The red signal time computed was entered in the ?Red_min? field manually, and is illustrated in Figure 4.1. 4.2.2 Unsignalized Intersection Delay The delay caused by the unsignalized intersections can be determined using equation 2.4 discussed in the literature review chapter is illustrated below as: ( )2 * 10.05 2n kd ?? ?= ?? ? ? ? (4.3) where: d2 = delay at an unsignalized roadway intersection (min), n = number of links ending at the intersection, k = number of links exiting from the intersection. N*(k -1) = the number of possible turns at the intersection. A new field named ?Unsigndel? was added to the attribute table of the GIS roadway network file. The information regarding the number of links entering and exiting at an unsignalized intersection can not be retrieved from the attribute table of GIS file. This information on the number of links entering and exiting the intersection were retrieved by manually observing each unsignalized intersection in the COA GIS map. Next, the delay at each unsignalized intersection was calculated using equation 4.3. The 79 information on the stop sign locations at intersections in the COA roadway network was not available. Therefore it was assumed that vehicle traveling on a major and minor arterial will not experience any delay at unsignalized intersections joining the major and minor arterials to local and collector roads. Therefore the unsignalized intersection delay values are assigned only to the local and collector roads that intersect with major and minor arterials. If at an unsignalized intersection is caused by major and minor arterials or two major arterials or two minor arterials, then an equal delay value is assigned to all the links at that intersection. Similarly when local and collector roads meet at an unsignalized intersection equal delay is assigned to all the links at the intersection. 4.2.3 Updating Speed Limit Data As mentioned earlier, there are many local streets that do not have a value for the corresponding speed limit in the COA roadway network attribute table. The equation used to calculate the travel time along the link will compute a travel time of infinity if the speed limit on the street is zero. It was observed that almost all of the local street speed limit values provided contained a value equal to zero. As mentioned in previous chapters, the majority of the local roads on which the speed limit value was missing are residential streets. An assumption was made determining that the minimum speed limit value for these streets will be equal to 25 mph. A field was added to the attribute table as ?Speedlim1?. The ?SPEEDLIMIT? field furnished current speed limit values for each street. A small VB script was created to replace all ?0 mph? values with ?25 mph? values. This script is presented in Table 4.4. Figure 4.1 provides an illustration of the 80 ?Speedlim1? field showing that a value of 25 mph was inserted as the speed limit for all records having null values in the ?SPEEDLIMIT? field. Table 4.4 VB Script for Speed Limit Update Code Dim speed as Integer Dim SPEEDLIMIT as Integer If [SPEEDLIMIT] = 0 Then speed = 25 Else speed = [SPEEDLIMIT] End If Value in Result Field Speed 4.2.4 Traffic Growth Factors The amount of traffic experienced on a roadway network will fluctuate year by year due to various reasons such as growth in population, increase in employment opportunities, and increases in number of vehicles in an area. Therefore traffic growth factors for the area under consideration need to be considered during model development. The GIS models being developed have to suit the traffic condition for the years the haul truck data is being collected (i.e. 2006). The AADT data for the COA is provided by the AADT GIS file. This file contains the AADT values for various location within the roadway network and were collected during a number of years ranging from 1996 to 2004. Of the 460 AADT values collected, a significant amount of AADT data was collected during the years 2000 (151 values) and 2004 (141 values). A few AADT data values were collected during the years 1996 (3 values) and 1998 (2 values) and no AADT data values were collected from 2000. Projecting the AADT data of the roadway network to the year 2006 will translate all AADT information to the current year. 81 The AADT values of each link are projected to the year 2006 using equation 4.4: AADT (2006) = AADT * (1 + AGR)n (4.4) where: AADT = available AADT value for a particular year AGR = annual growth rate n = difference in number of years between the year AADT data is collected to 2006 From previous data collection efforts an annual growth rate of 2% was considered for traffic projection purposes. Using the date and AADT data provided in the attribute table, AADT data was projected to 2006 and represented in the attribute table as ?PRO_AADDT? field. Table 4.5 furnishes the VB script used for projecting the AADT data to the year 2006. Figure 4.1 shows the attribute table after updating the AADT data to 2006. An assumption was made that a null value in the AADT data will return free flow speed conditions on the roadway at all times. 82 Table 4.5 VB Script for Projecting the AADT Data Code Dim a As Long Dim b As Integer Dim c As Double Dim e As Double b = 2006 ? [DATE_] If b = 0 Then c = 1 ElseIf b = 1 Then c = 1.02 ElseIf b = 2 Then c = 1.0404 ElseIf b = 3 Then c = 1.061208 ElseIf b = 4 Then c = 1.08243216 ElseIf b = 5 Then c = 1.1040808032 ElseIf b = 6 Then c = 1.126162419264 ElseIf b = 7 Then c = 1.14868566764928 ElseIf b = 8 Then c = 1.1716593810022656 ElseIf b = 9 Then c = 1.195092568622310912 ElseIf b = 10 Then c = 1.21899441999475713024 ElseIf b = 11 Then c = 1.2433743083946522728448 EndIf If [Integer] = 0 Then a = 0 Else e = c * [Integer] a = e EndIf Value in Result Field a 4.2.5 Network Travel Time The time required to traverse a street or a network link can be calculated using the volume delay function provided as Equation 2.1 which was discussed in the literature review is shown below as [Hedayat and Iravani, 1999]: 83 4 0( ) * * 1 0.15* * xt x l t c w ? ?? ?= + ? ?? ?? ?? ? ? ? (4.5) where: t(x) = travel time on the link (min), l = length of the link (ft), t0 = travel time for traveling unit length at free flow speed of the link (min), x = traffic on the link (veh/h), c = capacity of the link per lane (veh/h), and w = number of lanes in the link. In equation 4.5 the value of the denominator represents the capacity of the link per lane in one direction multiplied by the number of lanes on the link in the same direction. This represents the capacity of the roadway in one direction. The HCM furnishes service volume capacities of urban arterials as 800 veh/h for one lane, 1,620 veh/h for two lane, 2,430 veh/h for three lane, and 3,250 veh/h for four lane roadways. The GIS model was updated in Chapter 3 to incorporate the number of lanes of each street in the COA roadway network. An attribute field named ?Capacity? was created for furnishing service volume capacity for each road to the GIS models. The VB script furnished in Table 4.6 will position the capacity information in the attribute table of the GIS network accordingly. 84 Table 4.6 VB Script for furnishing Capacity of Roadway Code Dim a As Double If [Nooflane] = 1 Then a = 800 If [Nooflane] = 2 Then a = 1620 ElseIf [Nooflane] = 3 Then a = 1620 ElseIf [Nooflane] = 4 Then a = 3250 ElseIf [Nooflane] = 5 Then a = 3250 Else a = 800 EndIf Value in Result Field A The COA GIS road network is projected in the NAD 1983 StatePlane Alabama East FIPS 0101 Feet coordinate system. As mentioned earlier in the data collection section, this projection better represents the distance measurement of an actual map. Also the grade of roadway network does not affect the linear measurement of roadway link as it will be less than 10%. The length of each link has to be calculated in feet using the GIS map. A data field is created entitled ?lengthfeet? in the attribute table of the GIS file. The VB script presented in Table 4.7 was used to calculate get length information to the ?lengthfield? field of the attribute table. Table 4.7 VB Script for Getting Length of the Roadway Link Code Dim length as double Dim pCurve as Icurve Set pCurve = [shape] length = pCurve.Length Value in Result Field Length 85 4.3 TIME-DEPENDENT GIS MODELS The development of GIS models to account for fluctuations in traffic conditions at the time the haul operation occurs is considered in time-dependent GIS modeling. The total time required to complete the haul operation depends on the traffic conditions of the roadway network at the time of a haul operation. Therefore information about the date and time at which a haul operation takes place is essential for model development. The Global Positioning Systems (GPS) data collected provides the information about the date and time of each haul operation. The date and time information used to develop time-dependent GIS models included the month, weekday, and hour on which the haul operation occurred. Table 4.8 provides the date and time information of each relevant haul operation data collected by the GPS data logger and used for development of time- dependent GIS models. 86 Table 4.8 Date and Time Information Used for GIS Models Route Number Haul Type Time Interval of Haul Day of the Week Month of the Year Haul 11.00-12.00 P.M. Thursday May 1 Return 12.00-1.00 P.M. Thursday May Haul 12.00-1.00 P.M. Thursday May 2 Return 12.00-1.00 P.M. Thursday May Haul 2.00-3.00 P.M. Wednesday May 3 Return 3.00-4.00 P.M. Wednesday May Haul 11.00-12.00 P.M. Friday May 4 Return 11.00-12.00 P.M. Friday May Haul 12.00-1.00 P.M. Friday May 5 Return 1.00-2.00 P.M. Friday May Haul 7.00-8.00 A.M. Thursday June 6 Return 7.00-8.00 A.M. Thursday June Haul 8.00-9.00 A.M. Thursday June 7 Return 8.00-9.00 A.M. Thursday June Haul 10.00-11.00 A.M. Monday June 8 Return 11.00-12.00 P.M. Monday June Haul 10.00-11.00 A.M. Thursday June 9 Return 11.00-12.00 P.M. Thursday June Haul 12.00-1.00 P.M. Thursday June 10 Return 12.00-1.00 P.M. Thursday June Haul 3.00-4.00 P.M. Monday June 11 Return 4.00-5.00 P.M. Monday June Haul 4.00-5.00 P.M. Monday June 12 Return 5.00-6.00 P.M. Monday June Haul 3.00-4.00 P.M. Thursday June 13 Return 3.00-4.00 P.M. Thursday June Haul 7.00-8.00 P.M. Tuesday June 14 Return 7.00-8.00 P.M. Tuesday June Haul 8.00-9.00 A.M. Wednesday June 15 Return 9.00-10.00 A.M. Wednesday June Haul 9.00-10.00 A.M. Wednesday June 16 Return 10.00-11.00 A.M. Wednesday June Haul 2.00-3.00 P.M. Wednesday June 17 Return 3.00-4.00 P.M. Wednesday June Haul 3.00-4.00 P.M. Wednesday June 18 Return 4.00-5.00 P.M. Wednesday June 87 Table 4.8 Date and Time Information Used for GIS Models (Cont?d.) Route Number Haul Type Time Interval of Haul Day of the Week Month of the Year Haul 12.00-1.00 P.M. Friday June 19 Return 1.00-2.00 P.M. Friday June Haul 7.00-8.00 A.M. Friday June 20 Return 8.00-9.00 A.M. Friday June Haul 8.00-9.00 A.M. Friday June 21 Return 9.00-10.00 A.M. Friday June Haul 11.00-12.00 P.M. Friday June 22 Return 12.00-1.00 P.M. Friday June Haul 9.00-10.00 A.M. Friday June 23 Return 10.00-11.00 A.M. Friday June Haul 12.00-1.00 P.M. Monday June 24 Return 1.00-2.00 P.M. Monday June Haul 7.00-8.00 A.M. Friday June 25 Return 8.00-9.00 A.M. Friday June Haul 7.00-8.00 A.M. Friday June 26 Return 8.00-9.00 A.M. Friday June Haul 10.00-11.00 A.M. Friday June 27 Return 11.00-12.00 P.M. Friday June Haul 1.00-2.00 P.M. Monday June 28 Return 1.00-2.00 P.M. Monday June Haul 2.00-3.00 P.M. Monday June 29 Return 2.00-3.00 P.M. Monday June Haul 12.00-1.00 P.M. Tuesday June 30 Return 1.00-2.00 P.M. Tuesday June Haul 7.00-8.00 A.M. Tuesday June 31 Return 10.00-11.00 A.M. Tuesday June Haul 12.00-1.00 P.M. Tuesday June 32 Return - - - Haul 12.00-1.00 P.M. Wednesday June 33 Return 2.00-3.00 P.M. Wednesday June Haul 2.00-3.00 P.M. Wednesday June 34 Return 4.00-5.00 P.M. Wednesday June Haul 2.00-3.00 P.M. Tuesday June 35 Return - - - Haul 3.00-4.00 P.M. Thursday June 36 Return 4.00-5.00 P.M. Thursday June Note: * = Truck was stationary along route for non-traffic related reasons. To obtain the probable traffic conditions on the roadway network at the time of haul operation, the traffic multiplication factors developed for the COA were used. Equation 88 4.6 was used to determine the traffic conditions on the COA roadway network at the hour of haul operation: AADT = (Traffic Count) * (HF) * (DF) * (MF) (4.6) where, HF = hourly factor = total hourly percentage of traffic during given period of time 100 ? ? ? ? ? ? ? ? ? ?? ? DF = daily factor (weekday multiplication factor) MF = monthly factor AADT = Annual Average Daily Traffic Traffic Count = total traffic count of the street in both direction for a given period of time The traffic data collected by the COA does not provide information on the directional distribution of the traffic. Therefore it is assumed that the traffic distribution is equal to 50% for any street on the COA roadway work. A new field of named ?50ADT? was created in the attribute table of the GIS base model file. The traffic multiplication factors (monthly, daily and hourly), for the date and time of haul operation were extracted from Table 3.5, Table 3.6, and Table 3.7. The field calculator application is used to calculate the values at the ?50ADT? field in the attribute table. The formula line used in the field calculator is shown below in equation 4.7. A value of 2 is used in the denominator to account for the 50% directional distribution: ([PRO_AADDT] * [hourly percentage of traffic] 50ADT = ( [weekday multiplication factor] * [monthly factor] * 100 * 2) (4.7) 89 4.3.1 Calculating Signalized Intersection Delay A new field with a name ?signdelay? was created to furnish the signalized delay value for every link of the GIS models. Equation 2 now has all the information required to calculate signal delay in the attribute table of GIS file. Any vehicle traveling through an intersection will only travel on two links of the intersection. Therefore one-half of the calculated delay value at the intersection is assigned to each link joining the intersection. Table 4.9 below provides the VB script used to calculate the signal delay for the ?signdelay? field in the attribute table of GIS file. Table 4.9 VB Script for Calculating Delay at Signalized Intersection Code Dim a As Double Dim b As Double Dim c As Double Dim d As Double Dim e As Double Dim f As Double If [Cycle_leng] = 100 Then b = [r_square] c = [50ADT] d = 1800 ? c e = b * 60 * 1800 f = 2 * 100 * d * 2 a = e / f ElseIf [Cycle_leng] = 70 Then b = [r_square] c = [50ADT] d = 1700 ? c e = b * 60 * 1700 f = 2 * 70 * d * 2 a = e / f Else a = 0 EndIf Value in Result Field a 90 4.3.2 Calculating Network Travel Time The travel time of the links in the COA roadway network are calculated using equation 4.5 discussed in section 4.2.5. For calculating the travel time of the links, the speed limit value is assumed as free flow speed of the link. Now all the data required to calculate the travel time on the links of the COA GIS roadway network are available in the attribute table. A new field with ?traveltime? name is added to the attribute table of GIS file. Table 4.10 below presents the VB script used to calculate the total travel time on each roadway network link including the delay experienced at each intersection. Travel time values are recorded in the ?traveltime? field of the attribute table of GIS roadway network file. Table 4.10 VB Script for Travel Time Calculation Code Dim a As Double Dim b As Double Dim c As Double Dim d As Double Dim e As Double Dim f As Double Dim g As Double Dim h As Double Dim i As Double Dim l As Double Dim m As Double Dim n As Double Dim p As Double c = [Capacity] b = [50ADT] h = [lengthfeet] i = [Speedlim1] n = [signdelay] p = [Unsigndel] d = b * b * b * b e = c * c* c * c f = 0.15 * d / e g = f +1 l = 60 / i m = h * l * g / 5280 a = m + n+ p Value in Result Field a 91 4.4 TIME-DEPENDENT GIS MODEL VALIDATION The adequacy of the time-dependent model was tested against the actual data collected from the GPS data of haul truck operation. The actual time required to complete a haul or return route is compared against the same haul route modeled by using time-dependent GIS models. After calculating the travel time of the links in the attribute table, the actual haul route GPS points are overlaid on the time-dependent route the GIS model developed using the appropriate traffic conditions. Figure 4.2(a) shows the actual haul route GPS points overlaid on the modeled GIS road network in the background. The Network Analyst extension at ArcGISTM software was used to model the time-dependent GIS route. The time-dependent GIS route was made begin and end at the exact origin and destination of actual haul route. The model was then programmed to follow the exact route of the actual GPS data collected. Figure 4.2(b) shows the GIS model route exactly overlaid with the actual haul route GPS points on the modeled GIS road network. 92 (a) GIS Model Route on GPS Haul Route (b) Haul Route GPS points on GIS Model Developed Figure 4.2 Similarity of GIS Model of Actual Haul Route and GPS Haul Routes. The time predicted by the GIS model developed for the actual route is obtained from ?traveltime? field of the attribute table. The value obtained from the ?traveltime? field is in minutes. The time required for the actual haul and return route is obtained in seconds from the actual GPS data and it is converted into minutes for validation purpose. For the purpose of model validation the required travel time for the actual haul and return routes obtained in the field was compared with the travel time predicted for the same routes by Location of Batch Plant Ending Point of Haul * Starting Point of Haul Haul Route and its GPS points GIS Model Route Following the Haul Route N N N 93 the time-dependent GIS model. Table 4.11 provides the comparison of actual and modeled travel times for all haul and return routes. Table 4.11 Data Summary for GIS Model Validation ACTUAL ROUTE Travel Time Route Number Haul Type Distance (mile) Actual (min) Model (min) Difference (min) Percentage Difference in Actual and Model Times Haul 2.540 5.12 5.30 -0.18 -3.58 1 Return 2.540 5.02 5.00 0.02 0.33 Haul 2.540 4.72 5.38 -0.67 -14.13 2 Return 2.540 4.95 5.00 -0.05 -1.01 Haul 8.301 16.27 14.58 1.68 10.35 3 Return 8.301 13.87 14.15 -0.28 -2.04 Haul 4.760 12.98 11.37 1.62 12.45 4 Return 4.760 12.53 11.10 1.43 11.44 Haul 4.760 13.60 11.62 1.98 14.58 5 Return 4.760 12.32 11.27 1.05 8.53 Haul 4.528 10.33 9.87 0.47 4.52 6 Return 4.528 11.52 9.58 1.93 16.79 Haul 4.528 13.67 9.82 3.85 28.17 7 Return 4.528 13.65 9.55 4.10 30.04 Haul 4.379 11.95 9.77 2.18 18.27 8 Return 4.379 10.05 9.65 0.40 3.98 Haul 7.369 9.57 9.55 0.02 0.17 9 Return 7.699 11.23 10.88 0.35 3.12 Haul 3.219 6.42 5.58 0.83 12.99 10 Return 3.219 5.87 5.58 0.28 4.83 Haul 3.858 8.47 7.12 1.35 15.94 11 Return 3.858 8.22 7.12 1.10 13.39 Haul 3.858 7.87 7.12 0.75 9.53 12 Return 3.858 9.17 7.10 2.07 22.55 Haul 4.265 10.37 9.78 0.58 5.63 13 Return 4.471 10.82 10.83 -0.02 -0.15 Haul 4.753 10.87 10.17 0.70 6.44 14 Return 4.753 9.57 10.17 -0.60 -6.27 Haul 4.098 9.40 8.62 0.78 8.33 15 Return 4.098 7.80 8.72 -0.92 -11.75 Haul 4.098 8.72 8.72 0.00 0.00 16 Return 4.098 8.83 8.80 0.03 0.38 Haul 3.858 7.83 7.10 0.73 9.36 17 Return 3.858 8.67 7.12 1.55 17.88 Haul 3.398 6.50 6.97 -0.47 -7.18 18 Return 3.398 7.33 7.02 0.32 4.32 94 Table 4.11 Data Summary for GIS Model Validation (Cont?d.) ACTUAL ROUTE Travel Time Route Number Haul Type Distance (mile) Actual (min) Model (min) Difference (min) Percentage Difference in Actual and Model Times Haul 3.287 8.25 6.82 1.43 17.37 19 Return 3.363 6.88 7.00 -0.12 -1.69 Haul 3.990 8.27 8.45 -0.18 -2.22 20 Return 3.960 10.52 8.42 2.10 19.97 Haul 1.877 4.47 3.55 0.92 20.52 21 Return 1.877 3.68 3.78 -0.10 -2.71 Haul 2.800 5.87 6.02 -0.15 -2.56 22 Return 2.800 4.47 5.02 -0.55 -12.31 Haul 1.877 4.67 3.57 1.10 23.57 23 Return 1.877 3.40 3.78 -0.38 -11.27 Haul 6.384 12.08 11.77 0.32 2.62 24 Return 6.384 12.63 11.70 0.93 7.39 Haul 7.677 10.80 10.05 0.75 6.94 25 Return 8.111 11.10 10.93 0.17 1.50 Haul 3.490 6.73 6.03 0.70 10.40 26 Return 3.490 5.70 6.03 -0.33 -5.85 Haul 7.677 10.37 10.20 0.17 1.61 27 Return 8.111 11.73 11.63 0.10 0.85 Haul 2.800 5.77 6.02 -0.25 -4.34 28 Return 2.800 5.40 5.02 0.38 7.10 Haul 3.442 6.53 5.88 0.65 9.95 29 Return 3.442 6.13 5.88 0.25 4.08 Haul 3.490 7.43 6.07 1.37 18.39 30 Return 3.490 7.18 6.02 1.17 16.24 Haul 3.858 9.03 7.07 1.97 21.77 31 Return 3.858 7.60 7.08 0.52 6.80 Haul 7.671 11.93 11.85 0.08 0.70 32 Return* - - - - - Haul 4.399 11.63 10.32 1.32 11.32 33 Return 4.399 10.73 10.25 0.48 4.50 Haul 7.952 15.20 14.07 1.13 7.46 34 Return 7.952 15.87 14.12 1.75 11.03 Haul 4.760 12.70 11.38 1.32 10.37 35 Return* - - - - - Haul 7.952 13.97 14.12 -0.15 -1.07 36 Return 7.624 17.67 15.43 2.23 12.64 Cumulative Values 317.667 652.47 602.41 50.07 7.67 Average Values 4.540 9.32 8.61 0.72 7.62 Standard Deviation 1.830 3.31 2.93 0.38 11.48 Note : * = Truck was stationary along route for non-traffic related reasons. 95 From Table 4.11, the total distance of the haul truck data collected was 317.667 miles and trucks consumed 652.47 minutes to travel on that distance. The average travel distance and time of the actual routes was 4.54 miles and 9.32 minutes respectively and the averages of GIS model was 4.54 miles and 8.61 minutes respectively. A difference of 7.6% in travel time prediction is observed. The standard deviation of actual travel time (3.31 min) is a larger than standard deviation of model travel time (2.93 min). This result shows that the predicted travel time has less variation in travel times than actual travel times. Two graphs were generated computing actual travel time vs. predicted travel time obtain a better understanding and representation of the time-dependent GIS model results. Separate graphs are produced for haul and return routes. Figure 4.3 shows the graph of actual travel times vs. predicted travel times for haul routes and Figure 4.4 illustrates travel times for the return routes. It was observed that each model provides good prediction for the travel time of haul and return routes. A regression line for R2 value 1.0 is drawn for each graph. As the travel time is a null value for the zero distance, the R2 value and regression equation of the dataset is found by forcing the regression line through origin. The travel time prediction has an R2 value of 0.9198 and regression equation y = 0.9111x for the haul route models and R2 value of 0.9082 and regression equation y = 0.9193x for the return route models. This proves the reliability of GIS time-dependent models with respect to predicting travel time in the dynamic roadway network. These results establish the credibility that the routes found based on the GIS time-dependent models represent the actual roadway network conditions. 96 y = 0.9111x R2 = 0.9198 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Actual Travel Time (minutes) Pre dic ted Tr av el Ti me (m inu tes ) Figure 4.3 GIS Model Validation for Actual vs. Predicted Travel Time of Haul Routes. 97 y = 0.9193x R2 = 0.9082 0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20 Actual Travel Time (minutes) Pre dic ted Tr av el Ti me (m inu tes ) Figure 4.4 GIS Model Validation for Actual vs. Predicted Travel Time of Return Routes. 4.5 OPTIMIZATION OF HAUL ROUTES The haul routes can be optimized by traveling on routes which will either reduce save travel time or travel distance, or both. This can be achieved by finding better routes for each actual haul routes using the time-dependent GIS models developed for each 98 routes. The GIS models developed in the previous sections for actual haul route timings provide excellent prediction of the particular haul route and good representation of the dynamic conditions of the COA roadway network at the time of haul operation. The Network Analysts extension of ArcGISTM uses Dijkstra algorithm to find the shortest route. It will calculate the shortest route based on any one parameter assigned to it. The shortest distance and shortest time models can be found by assigning length and time parameters to the Network Analyst. As the GIS modeling is carried out for truck travel, care should be taken that a large portion of the route should pass through major and minor arterials in the roadway network. This goal is achieved by not permitting the haul routes to travel on the local and collector streets till the route in close proximity to the planned destination. The barrier function in the Network Analyst extension is used to accomplish this. The barrier function temporarily deactivates links in the GIS roadway network to prevent the route from traversing the identified link (i.e. residential street) In determining the route for the haul route operation, the origin (i.e. plant location) and destination (i.e. delivery location) are known. In the GIS model, the origin and destination locations are found visually and marked on the GIS network using the Network Analyst extension. The route selection is then solved by the Network Analyst, and the GIS map will display a shortest time and distance route from the origin to destination. The GIS models were also solved for the timings of actual haul or return routes. To accomplish this, the model was forced to travel on the actual haul routes collected by GPS. For this purpose, the barrier function of the Network Analysts was used. Figure 4.5(a) shows the GIS modeled route of actual haul route data. To find the 99 shortest distance route the distance attribute is assigned to the Network Analyst. Then the GIS model route for the shortest distance is found using the GIS time-dependent model developed. While finding the shortest route on the roadway network, if the route travels on the local or collector roads during initial phase of the haul or return route, then a barrier will be placed at the point where the route enters the collector or local street from a major or minor arterial. Figure 4.5(b) shows the shortest distance path for the same origin to destination location. Similarly, to find the shortest time routes for the haul operation, the travel time attribute is assigned to the Network Analysts. In this case, the GIS modeled route for the shortest time is instigated using the GIS time-dependent model developed earlier for the required haul operation. The barrier functions were also used to limit the shortest time route from using local residential streets in the roadway network. Figure 4.5(c) depicts the shortest time route for the same origin to destination locations. Figure 4.5(d) illustrates the superimposition of all three routes on the COA GIS map. Similarly shortest distance and time routes are found for each haul operation using their corresponding time-dependent GIS model. All the shortest distance and time models developed for each of haul routes have been represented in tabular format in Table 4.12. 100 (a) Actual Haul Route (b) Shortest Distance Route (c) Shortest Time Route (d) All Three Routes Figure 4.5 Actual and Modeled GIS Routes. N N N N Table 4.12 Comparison of Actual Route and Modeled Routes ACTUAL ROUTE SHORTEST DISTANCE MODEL SHORTEST TIME MODEL Difference in Difference in Route Number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 2.540 5.12 2.540 5.30 0.000 -0.183 2.540 5.30 0.000 -0.18 1 Return 2.540 5.02 2.540 5.00 0.000 0.017 2.540 5.00 0.000 0.02 Haul 2.540 4.72 2.540 5.38 0.000 -0.667 2.540 5.38 0.000 -0.67 2 Return 2.540 4.95 2.540 5.00 0.000 -0.050 2.540 5.00 0.000 -0.05 Haul 8.301 16.27 8.301 14.58 0.000 1.683 8.301 14.58 0.000 1.68 3 Return 8.301 13.87 8.301 14.15 0.000 -0.283 8.301 14.15 0.000 -0.28 Haul 4.760 12.98 4.700 11.68 0.060 1.300 4.751 11.02 0.009 1.97 4 Return 4.760 12.53 4.700 11.42 0.060 1.117 4.751 10.75 0.009 1.78 Haul 4.760 13.60 4.700 11.82 0.060 1.783 4.751 11.18 0.009 2.42 5 Return 4.760 12.32 4.700 11.50 0.060 0.817 4.751 10.87 0.009 1.45 Haul 4.528 10.33 4.518 9.42 0.009 0.917 4.518 9.42 0.009 0.92 6 Return 4.528 11.52 4.518 9.15 0.009 2.367 4.518 9.15 0.009 2.37 Haul 4.528 13.67 4.518 9.38 0.009 4.283 4.518 9.38 0.009 4.28 7 Return 4.528 13.65 4.518 9.12 0.009 4.533 4.518 9.12 0.009 4.53 Haul 4.379 11.95 4.379 9.53 0.000 2.417 4.370 9.53 0.009 2.42 8 Return 4.379 10.05 4.379 9.38 0.000 0.667 4.370 9.38 0.009 0.67 Haul 7.369 9.57 5.800 11.52 1.569 -1.950 7.369 9.55 0.000 0.02 9 Return 7.699 11.23 5.800 11.80 1.899 -0.567 7.699 10.88 0.000 0.35 Haul 3.219 6.42 3.219 5.58 0.000 0.833 3.219 5.58 0.000 0.83 10 Return 3.219 5.87 3.219 5.58 0.000 0.283 3.219 5.58 0.000 0.28 Haul 3.858 8.47 3.588 6.83 0.270 1.633 3.589 6.58 0.269 1.88 11 Return 3.858 8.22 3.588 6.83 0.270 1.383 3.589 6.58 0.269 1.63 Haul 3.858 7.87 3.588 6.83 0.270 1.033 3.589 6.60 0.269 1.27 12 Return 3.858 9.17 3.588 6.83 0.270 2.333 3.589 6.58 0.269 2.58 10 1 Table 4.12 Comparison of Actual Route and Modeled Routes (Cont?d.) ACTUAL ROUTE SHORTEST DISTANCE MODEL SHORTEST TIME MODEL Difference in Difference in Route Number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 4.265 10.37 4.265 9.78 0.000 0.583 4.265 9.78 0.000 0.58 13 Return 4.471 10.82 4.464 10.30 0.007 0.517 4.464 10.30 0.007 0.52 Haul 4.753 10.87 4.753 10.17 0.000 0.700 4.753 10.17 0.000 0.70 14 Return 4.753 9.57 4.753 10.17 0.000 -0.600 4.753 10.17 0.000 -0.60 Haul 4.098 9.40 4.098 8.62 0.000 0.783 4.098 8.62 0.000 0.78 15 Return 4.098 7.80 4.098 8.72 0.000 -0.917 4.098 8.72 0.000 -0.92 Haul 4.098 8.72 4.098 8.72 0.000 0.000 4.098 8.62 0.000 0.10 16 Return 4.098 8.83 4.098 8.80 0.000 0.033 4.098 8.80 0.000 0.03 Haul 3.858 7.83 3.588 6.83 0.270 1.000 3.589 6.58 0.269 1.25 17 Return 3.858 8.67 3.588 6.83 0.270 1.833 3.589 6.60 0.269 2.07 Haul 3.398 6.50 3.398 6.97 0.000 -0.467 3.398 6.97 0.000 -0.47 18 Return 3.398 7.33 3.398 7.02 0.000 0.317 3.398 7.02 0.000 0.32 Haul 3.287 8.25 3.287 6.82 0.000 1.433 3.287 6.82 0.000 1.43 19 Return 3.363 6.88 3.363 7.00 0.000 -0.117 3.378 6.68 -0.015 0.20 Haul 3.990 8.27 3.990 8.45 0.000 -0.183 3.990 8.45 0.000 -0.18 20 Return 3.960 10.52 3.960 8.42 0.000 2.100 3.990 8.42 -0.030 2.10 Haul 1.877 4.47 1.877 3.55 0.000 0.917 1.877 3.55 0.000 0.92 21 Return 1.877 3.68 1.877 3.78 0.000 -0.100 1.877 3.78 0.000 -0.10 Haul 2.800 5.87 2.800 6.02 0.000 -0.150 2.800 6.02 0.000 -0.15 22 Return 2.800 4.47 2.800 5.02 0.000 -0.550 2.800 5.02 0.000 -0.55 Haul 1.877 4.67 1.877 3.57 0.000 1.100 1.877 3.57 0.000 1.10 23 Return 1.877 3.40 1.877 3.78 0.000 -0.383 1.877 3.78 0.000 -0.38 Haul 6.384 12.08 6.046 12.62 0.338 -0.533 6.384 11.77 0.000 0.32 24 Return 6.384 12.63 6.046 12.33 0.338 0.300 6.384 11.70 0.000 0.93 10 2 Table 4.12 Comparison of Actual Route and Modeled Routes (Cont?d.) ACTUAL ROUTE SHORTEST DISTANCE MODEL SHORTEST TIME MODEL Difference in Difference in Route Number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 7.677 10.80 7.099 14.72 0.577 -3.917 7.677 10.05 0.000 0.75 25 Return 8.111 11.10 7.099 14.60 1.012 -3.500 8.111 10.93 0.000 0.17 Haul 3.490 6.73 3.490 6.03 0.000 0.700 3.490 6.03 0.000 0.70 26 Return 3.490 5.70 3.490 6.03 0.000 -0.333 3.490 6.03 0.000 -0.33 Haul 7.677 10.37 7.099 14.85 0.577 -4.483 7.677 10.20 0.000 0.17 27 Return 8.111 11.73 7.099 15.17 1.012 -3.433 7.677 11.63 0.434 0.10 Haul 2.800 5.77 2.800 6.02 0.000 -0.250 2.800 6.02 0.000 -0.25 28 Return 2.800 5.40 2.800 5.02 0.000 0.383 2.800 5.02 0.000 0.38 Haul 3.442 6.53 3.442 5.88 0.000 0.650 3.442 5.88 0.000 0.65 29 Return 3.442 6.13 3.442 5.88 0.000 0.250 3.442 5.88 0.000 0.25 Haul 3.490 7.43 3.490 6.07 0.000 1.367 3.490 6.07 0.000 1.37 30 Return 3.490 7.18 3.490 6.02 0.000 1.167 3.490 6.02 0.000 1.17 Haul 3.858 9.03 3.588 6.82 0.270 2.217 3.589 6.57 0.269 2.47 31 Return 3.858 7.60 3.588 6.82 0.270 0.783 3.589 6.58 0.269 1.02 Haul 7.671 11.93 7.671 11.85 0.000 0.083 7.671 11.85 0.000 0.08 32 Return* - - - - - - - - - - Haul 4.399 11.63 4.390 9.93 0.009 1.700 4.390 9.93 0.009 1.70 33 Return 4.399 10.73 4.390 9.90 0.009 0.833 4.390 9.90 0.009 0.83 Haul 7.952 15.20 7.615 14.50 0.338 0.700 7.952 14.07 0.000 1.13 34 Return 7.952 15.87 7.615 14.65 0.338 1.217 7.952 14.12 0.000 1.75 Haul 4.760 12.70 4.672 11.68 0.088 1.017 4.751 11.02 0.009 1.68 35 Return* - - - - - - - - - - Haul 7.952 13.97 7.615 14.65 0.338 -0.683 7.952 14.12 0.000 -0.15 36 Return 7.624 17.67 7.615 14.72 0.009 2.950 7.952 14.22 -0.328 3.45 Cumulative Values 317.668 652.43 306.779 619.70 10.889 32.73 315.334 591.18 2.334 61.25 Average Values 4.540 9.32 4.383 8.85 0.156 0.47 4.505 8.45 0.033 0.88 Standard Deviation 1.828 3.31 1.657 3.33 0.346 1.55 1.838 2.88 0.108 1.12 Note : * = Truck was stationary along route for non-traffic related reasons. 10 3 104 The optimization of haul routes by using shortest distance and shortest time routes resulted in savings of time and distance for the haul operations. The cumulative actual distance collected for the all haul operations is 317.668 miles and the total time consumed for it is 652.43 minutes. The average travel distance and time of the actual routes was 4.54 mile and 9.32 minutes respectively and the average values shortest distance model was 4.383 mile and 8.61 minutes respectively. The average distance traveled by the shortest time model is 4.505 miles with average travel time of 8.45 minutes. With the use of shortest distance GIS models routes, a saving of 10.9 miles is experienced with a time savings of 32.7 minutes in comparison to the actual routes traveled. For all routes the shortest time GIS models provided, a savings of 61.2 minutes and 2.3 miles was achieved over actual routes traveled. Only a little difference was observed in travel time and distance standard deviation values obtained for actual routes and shortest distance routes. But the standard deviation of actual travel time (3.31 min) is higher than the standard deviation of shortest time model travel time (2.88 min). This result shows that there was same type of variation in actual travel distance prediction and less variation in travel time prediction by the shortest time model. It has been observed that some routes selected by the shortest distance model and shortest time model are same as the actual route. Therefore to have better estimation of savings in terms time and distance by the GIS models, separate tables for comparing the modeled routes which provided alternative routes in comparison to actual routes taken have been developed. Table 4.13 compares the actual route with the shortest distance model routes for only the routes which are different from actual routes. 105 Table 4.13 Alternative GIS Modeled Shortest Distance Routes Compared Against Actual Routes Taken ACTUAL ROUTE SHORTEST ROUTE MODEL Difference in Route Number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 4.760 12.98 4.700 11.68 0.060 1.30 4 Return 4.760 12.53 4.700 11.42 0.060 1.12 Haul 4.760 13.60 4.700 11.82 0.060 1.78 5 Return 4.760 12.32 4.700 11.50 0.060 0.82 Haul 4.528 10.33 4.518 9.42 0.009 0.92 6 Return 4.528 11.52 4.518 9.15 0.009 2.37 Haul 4.528 13.67 4.518 9.38 0.009 4.28 7 Return 4.528 13.65 4.518 9.12 0.009 4.53 Haul 7.369 9.57 5.800 11.52 1.569 -1.95 9 Return 7.699 11.23 5.800 11.80 1.899 -0.57 Haul 3.858 8.47 3.588 6.83 0.270 1.63 11 Return 3.858 8.22 3.588 6.83 0.270 1.38 Haul 3.858 7.87 3.588 6.83 0.270 1.03 12 Return 3.858 9.17 3.588 6.83 0.270 2.33 Haul* - - - - - - 13 Return 4.471 10.82 4.464 10.30 0.007 0.52 Haul 3.858 7.83 3.588 6.83 0.270 1.00 17 Return 3.858 8.67 3.588 6.83 0.270 1.83 Haul 6.384 12.08 6.046 12.62 0.338 -0.53 24 Return 6.384 12.63 6.046 12.33 0.338 0.30 Haul 7.677 10.80 7.099 14.72 0.577 -3.92 25 Return 8.111 11.10 7.099 14.60 1.012 -3.50 Haul 7.677 10.37 7.099 14.85 0.577 -4.48 27 Return 8.111 11.73 7.099 15.17 1.012 -3.43 Haul 3.858 9.03 3.588 6.82 0.270 2.22 31 Return 3.858 7.60 3.588 6.82 0.270 0.78 Haul 4.399 11.63 4.390 9.93 0.009 1.70 33 Return 4.399 10.73 4.390 9.90 0.009 0.83 Haul 7.952 15.20 7.615 14.50 0.338 0.70 34 Return 7.952 15.87 7.615 14.65 0.338 1.22 Haul 4.760 12.70 4.672 11.68 0.088 1.02 35 Return* - - - - - - Haul 7.952 13.97 7.615 14.65 0.338 -0.68 36 Return 7.624 17.67 7.615 14.72 0.009 2.95 Cumulative values 176.931 365.55 166.042 346.05 10.889 19.50 Average Values 5.529 11.42 5.189 10.81 0.340 0.61 Standard Deviation 1.676 2.46 1.456 2.96 0.449 2.13 Note : * = Truck was stationary along route for non-traffic related reasons or the GIS model route was same as actual truck route 106 The optimization of haul routes by using shortest distance routes resulted in savings of time and distance for the haul operations. The shortest distance model found 32 alternative routes in comparison to the actual routes. The cumulative actual distance collected for the haul routes which were different from shortest distance routes is 176.931 miles and the total time consumed for it is 365.55 minutes. The average travel distance and time of the actual routes was 5.559 miles and 11.42 minutes respectively and the average values of the shortest distance model was 5.189 miles and 10.81 minutes respectively. With the use of shortest distance GIS model routes which are different from actual route, a saving of 10.889 miles (6.2%) is experienced with a time savings of 19.5 minutes (5.3%) in comparison to the actual routes traveled. The standard deviation values for the actual travel time was less than travel time prediction by shortest distance models. But the standard deviation values of travel distance was more in the case of actual routes in comparison with shortest distance route travel distance. This proves that the shortest distance model has less variation in predicting the travel distance when compared to travel time prediction. As shortest distance model is used for finding the shortest distance routes the variation is acceptable. Similarly, Table 4.14 compares the actual route with the shortest time model routes for the modeled routes which are different from actual routes taken. 107 Table 4.14 Alternative GIS Modeled Shortest Time Routes Compared Against Actual Routes Taken ACTUAL ROUTE SHORTEST ROUTE MODEL Difference in Route Number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 4.760 12.98 4.751 11.02 0.009 1.97 4 Return 4.760 12.53 4.751 10.75 0.009 1.78 Haul 4.760 13.60 4.751 11.18 0.009 2.42 5 Return 4.760 12.32 4.751 10.87 0.009 1.45 Haul 4.528 10.33 4.518 9.42 0.009 0.92 6 Return 4.528 11.52 4.518 9.15 0.009 2.37 Haul 4.528 13.67 4.518 9.38 0.009 4.28 7 Return 4.528 13.65 4.518 9.12 0.009 4.53 Haul 4.379 11.95 4.370 9.53 0.009 2.42 8 Return 4.379 10.05 4.370 9.38 0.009 0.67 Haul 3.858 8.47 3.589 6.58 0.269 1.88 11 Return 3.858 8.22 3.589 6.58 0.269 1.63 Haul 3.858 7.87 3.589 6.60 0.269 1.27 12 Return 3.858 9.17 3.589 6.58 0.269 2.58 Haul* - - - - - - 13 Return 4.471 10.82 4.464 10.30 0.007 0.52 Haul 3.858 7.83 3.589 6.58 0.269 1.25 17 Return 3.858 8.67 3.589 6.60 0.269 2.07 Haul* - - - - - - 19 Return 3.363 6.88 3.378 6.68 -0.015 0.20 Haul* - - - - - - 20 Return 3.960 10.52 3.990 8.42 -0.030 2.10 Haul* - - - - - - 27 Return 8.111 11.73 7.677 11.63 0.434 0.10 Haul 3.858 9.03 3.589 6.57 0.269 2.47 31 Return 3.858 7.60 3.589 6.58 0.269 1.02 Haul 4.399 11.63 4.390 9.93 0.009 1.70 33 Return 4.399 10.73 4.390 9.90 0.009 0.83 Haul 4.760 12.70 4.751 11.02 0.009 1.68 35 Return* - - - - - - Haul* - - - - - - 36 Return 7.624 17.67 7.952 14.22 -0.328 3.45 Cumulative values 117.855 282.13 115.522 234.58 2.336 47.55 Average Values 4.533 10.85 4.443 9.02 0.090 1.83 Standard Deviation 1.059 2.49 1.104 2.10 0.163 1.10 Note : * = Truck was stationary along route for non-traffic related reasons or the GIS model route was same as actual truck route 108 Here also optimization of haul routes by using shortest time routes resulted in savings of time and distance for the haul operations. Although the savings obtained in distance by shortest route model was infinitesimal. The shortest distance model found 26 routes alternative routes from actual routes. The cumulative actual distance collected for the haul routes which different from shortest distance routes is 117.855 miles and the total time consumed for it is 282.13 minutes. The average travel distance and time of the actual routes was 4.533 miles and 10.85 minutes respectively and the average values of the shortest time model was 4.443 miles and 9.02 minutes respectively. With the use of shortest time GIS model routes which are different from actual route, a saving of 2.336 miles (2.0%) is experienced with a time savings of 47.55 minutes (16.8%) in comparison to the actual routes traveled. The standard deviation values for the actual travel distance was less than travel distance prediction by shortest time models. But standard deviation values of travel time was more in the case of actual routes in comparison with shortest time route travel time. This proves that the shortest time model has less variation in predicting the travel time as compared to travel distance prediction. Here also the shortest time model is used for finding the shortest time routes, therefore the variation is acceptable. Table 4.15 summarizes the savings experienced by GIS model models in which model routes are different from actual routes. It is observed that a savings of 1.83 minutes (16.9%) in travel time is observed for average route using shortest time model routes and 0.340 miles (6.2%) savings in travel distance was observed by using shortest distance model. 109 Table 4.15 Savings Experienced by GIS Model Models Shortest Distance Model (n = 32 Routes) Shortest Time Model (n = 26 Routes) Actual Route Model Route Total Savings Percentage Savings Actual Route Model Route Total Savings Percentage Savings Total Distance (miles) 176.931 166.042 10.899 6.2 117.855 115.522 2.336 2.0 Total Time (minutes) 365.55 346.05 19.50 5.3 282.13 234.58 47.55 16.9 Average Distance (miles) 5.529 5.189 0.340 6.2 4.533 4.443 0.090 2.0 Average Time (Minutes) 11.42 10.81 0.61 5.3 10.85 9.02 1.83 16.9 Note : n = Number of routes The developed optimization technique will provide the truck drivers an opportunity to select between the shortest distance or the shortest time route depending on their specific requirements. All the data for the trucking operation is collected over a period of more than 15 days for 2 haul trucks. It was observed that the trucks made a maximum of 4 hauls per day. The truck drivers were given incentives if they could make more than 4 haul operation per day. The incentives were used to encourage drivers to save time in hopes of improving delivery production. The optimization technique offers a saving of nearly 17% of haul time, and will help drivers achieve more haul routes per day. If 8 hours of haul truck operation is involved per day for the batch plant, using the shortest time model, optimized for the route selection could save nearly an hour of batch plant operation, daily per truck. This scale of savings in time per day could bring a substantial amount of additional profit to the batch plant and will be especially helpful during the busy operation days. The travel distance saved by shortest distance models in terms of 110 will have substantial financial benefits in the long run. More importantly both shortest distance and time models provide savings in distance as well as time. 4.6 CONCLUSION The base GIS model was developed using all the network parameters common to each model. The time-dependent GIS models were developed for varying traffic conditions of COA roadway network. The inclusion of traffic multiplication factors are used to incorporate the variation of traffic with respect to time. The time-dependent models properly depicted the COA roadway network and found reliable travel time predictions. The travel time reliability of a haul operation will enable the batch plant to better estimate the total concrete truck cycle time. The reliable travel time prediction will assist in development of better vehicle dispatch optimizing methods. The optimization of the haul routes by the shortest distance and shortest time routes provided adequate results by saving time and distance for all haul types. 111 CHAPTER FIVE CONCLUSIONS AND RECOMMENDATIONS 5 5.1 CONTRIBUTIONS OF THE RESEARCH This research collects Geographic Information Systems (GIS) and traffic related data for the City of Auburn (COA) for the development of a base GIS roadway network model. The data collected for the COA includes roadway network data, Annual Average Daily Traffic (AADT) data, geo referenced arial images, and traffic signal locations. Additional traffic data over the past ten years for the COA was also collected. This additional data includes annual hourly traffic data which was collected for the three most recent years from an Automatic Traffic Recorder (ATR) count station located in the COA to be utilized for the development of time dependent traffic models. Global Positioning Systems (GPS) data of a haul truck operation is collected from two concrete trucks delivering ready-mix concrete to local contractors operating in the COA. The GPS data was collected using GPS instruments mounted on the haul trucks. The GPS data provided spatial and time information for the actual route traveled by the trucks during the haul operations. The GPS data also includes origin, destination, and route 112 selection information regarding the route traveled during the haul operation. There were 36 up-haul and 34 return-haul routes collected for model validation purposes. Once the GPS data collection was complete, time-dependent GIS models were developed to accurately reflect the traffic conditions experienced by the haul trucks during the time the haul took place. To accomplish this, traffic multiplication factors for the COA were developed using the hourly traffic data collected for an entire year. The percentage of traffic growth experienced by the COA was developed using additional traffic data collected over the past ten years. The AADT data and traffic signal information were incorporated into the base GIS COA roadway network model. A method to also incorporate traffic growth and traffic multiplication factors into the COA GIS roadway network is recognized. The travel time required for the haul operations within the COA network was predicted by calculating the travel time required to traverse roadway links and to overcome delay experienced at signalized and unsignalized intersection on a time- dependent basis. Therefore separate time-dependent GIS models were developed for each haul operation, therefore a method to develop the time-dependent models was established. The time-dependent GIS models developed exhibited very good correlation with the actual haul travel times recorded and collected by the GPS instruments. The reliability of travel time provided by the GIS modeled haul operation will bring value to the batch plant operation and good customer rapport with the contractor by providing on-time delivery of goods. The reliability of travel time 113 prediction will aid for efficient modeling of haul vehicles and optimization of dispatch scheduling. The network analyst extension of ArcGISTM uses Dijkstra algorithm to find the shortest route in a network. Shortest distance and shortest time were two types of optimization methods used for finding the better route in terms of distance and time from actual haul routes. It is observed that the optimization technique used to find the shortest distance and shortest time models for selected routes of the haul operations were effective. These models offered savings in both time and distance when compared to the actual routes selected by the haul trucks. The shortest time models realized 9.4% savings in travel time and the shortest distance model realized 3.5% savings in travel distance. The shortest time model will save in total amount of batch plant operation in a day. The effect by savings in time will be more pronounced in larger urban areas in comparison to the COA, where more options for route selection are available. The 3.5% savings derived by using the shortest distance model will give considerable financial benefits in the long run for batch plants in terms of reduced fuel costs. 5.2 SUGGESTIONS FOR FUTURE RESEARCH The time-dependent GIS models can be developed using more facets of roadway network data which affects the flow of traffic. According to Highway Capacity Manual (HCM) guidelines for determined the base free flow speed of the street will be affected by a number of parameters. Some of the important parameters according to HCM are the directional distribution of traffic, percentage of heavy vehicles in traffic, median 114 availability, width of the median, type of median, and pavement condition. A street may experience delay in one direction because of high directional distribution of traffic although total traffic on the street is under the roadway capacity. A higher number of heavy vehicles on the roadway will decrease the speed and capacity of the roadway in terms of number of vehicles. Presence, type, and a minimum width of median will help in assuming the base free flow speed of a street. The pavement condition on the street will affect the speed of the vehicles traveling on the street. Therefore, the data on all the aforementioned road network parameters can be collected for developing a GIS road network model that yields both results. Stop signs used for traffic control causes delay for the traffic on the street. Therefore the location of stop signs will provide greater detail for better GIS modeling. Additionally separate GIS files that contain future temporary traffic control zones and lane closure information would help to accurately model the existing road network conditions. This information on the construction, reconstruction, or maintenance operations on a street may reduce the capacity of the street or completely close the street to traffic. These additional data sets will help to more accurately represent the current traffic condition of the roadway network for purposes of modeling. Similar research can be extended to other vehicles such as passenger cars. This will require GPS data to be collected from passenger cars to validate the accuracy of the GIS models being used for the route selection. The described method of the development of time-dependent models for varying traffic condition may aid in the time prediction accuracy for Internet based route finding tools such as MapQuestTM and GoogleTM Maps. The researchers considered comparing popular MapQuestTM and GoogleTM Maps routes to GIS model routes in order to verify 115 the capabilities of the model and to ascertain the contribution of this research to the mainstream applications. The actual haul routes collected from GPS data are compared with the MapQuestTM and GoogleTM Maps routes and the results are explained in Appendix A. Appendix A also provides a comparison between the shortest haul route found by the time-dependent GIS models Vs. MapQuestTM and GoogleTM Maps. The data on the cost of haul truck operations, including the fuel cost and other operating overhead cost is not available in literature. A survey could be conducted to determine the cost associated with haul truck operations. This survey would be very helpful in quantifying the benefits of the GIS model in terms of cost savings. Data on the characteristics of truck engine performance could also be collected to accurately measure fuel usage for the development of accurate, comprehensive trucking costs. GPS data should also be collected for tracking the truck and identifying travel speeds while collecting truck engine performance data to develop correlations between operating variables. The grade of the street could also be incorporated in the GIS model as the truck performance is changes significantly while traveling on grades. The evaluation of truck performance will help to accurately model haul route selection on a GIS roadway network to optimized route selection based on minimizing fuel usage. 116 REFERENCES 1. Alabama State. Alabama Department Of Transportation Alabama (ALDOT). Traffic Statistics. http://aldotgis.dot.state.al.us/trafficvolume/viewer.htm. (November, 26 2006) 2. Alabama State. Alabama Department Of Transportation (ALDOT). Traffic Polling Data system. http://aldotgis.dot.state.al.us/traffic_counts/. (May 26, 2006) 3. American Association of State Highway and Transportation Officials (AASHTO). A policy on geometric design of highways and streets : 2004. Corp Author(s): American Association of State Highway and Transportation Officials. Washington: D.C., 2004. 4. Batley, R. Marginal valuations of travel time and scheduling, and the reliability premium. Transportation Research Part E: Logistics and Transportation Review. 2006 5. Cooksey, D. Understanding the Global Position System (GPS) Montana State University-Bozeman (MSUB) http://www.montana.edu/places/gps/understd.html. (July 26, 2006) 6. Dean, J. D. State Plane Coordinate System. Colorado State University. http://www.warnercnr.colostate.edu/class_info/nr502/lg3/datums_coordinates/spcs.html . (May 29, 2006) 7. Dion, F., Rakha, H., and Kang, Y. Comparison of delay estimates at under- saturated and over-saturated pre-timed signalized intersections. Transportation Research. Part B, Methodological. Vol. 38B, no. 2 (Feb. 2004), p. 99-122. ISSN: 0191- 2615 8. Eglese, R., Maden, W., and Slater, A. A road timetableTM to aid vehicle routing and scheduling. Computers & Operations Research. 33 (2006) 3508-3519 9. Environmental Systems Research Institute, Inc (ESRI). Data Types and Models. http://www.gis.com/implementing_gis/data/data_types.html. (May 29, 2006a) 10. Environmental Systems Research Institute, Inc (ESRI). Mean Sea Level, GPS, and the Geoid, http://www.esri.com/news/arcuser/0703/geoid1of3.html. (May 29, 2006b) 117 11. Environmental Systems Research Institute, Inc (ESRI). GIS Data. http://www.gis.com/implementing_gis/data/index.html. (May 29, 2006c) 12. Environmental Systems Research Institute, Inc (ESRI). Using Data in GIS. http://www.gis.com/implementing_gis/data/usingdata.html. (May 29, 2006d) 13. Environmental Systems Research Institute, Inc (ESRI). What is GIS? http://www.gis.com/whatisgis/index.html. (May 29, 2006e) 14. Environmental Systems Research Institute, Inc (ESRI). Why use GIS? http://www.gis.com/whatisgis/whyusegis.html. (May 29, 2006f) 15. Federal Highway Administration (FHWA). Congestion Mitigation, April 2006. http://www.ops.fhwa.dot.gov/congestionmitigation/congestionmitigation.htm (May 29, 2006a) 16. Federal Highway Administration (FHWA). Focus on Congestion. http://www.fhwa.dot.gov/congestion/ (May 29, 2006b) 17. Federal Highway Administration (FHWA). Manual On Uniform Traffic Control Devices For Streets And Highways (2003 Edition). Corp Author(s): United States.; Federal Highway Administration, U.S. Dept. of Transportation. 2003 18. Fleischmann, B., Gnutzmann, S., and Sandvo?, E. Dynamic Vehicle Routing Based on Online Traffic Information. Transportation Science. Vol. 38, No. 4, November 2004, pp. 420?433 19. Freundschuh, S. M., Smelser, J., and Neil, P, B. Wayfinding and Navigation Behavior. In International Encyclopedia of the Social &Behavioral Sciences, 16391-94. Oxford: Pergamon, 2001 20. GPS Basics, Montana State University-Bozeman (MSUB). http://www.montana.edu/places/gps/slides/GPSbasics_files/v3_document.htm. (November 26, 2006) 21. Hedayat, Z. A., and Iravani, H. I. Use of intersection delay functions to improve Reliability of Traffic assignment models. Presentation at the Ann. International EMME/2 Conference, Chicago: IL, 1999 22. Ichoua, S., Gendreau, M., Potvin, J. Vehicle dispatching with time-dependent travel times. European Journal of Operational Research. 144(2). (2003): 379-396. 23. Kakooza, R., Luboobi, L.S., and Mugisha, J.Y.T. Modeling Traffic Flow and Management at Un-signalized, Signalized and Roundabout Road Intersections. Journal of Mathematics and Statistics. 2005. 194-202 118 24. Kim, S., Lewis, M, E., and White, C. C. III,. Optimal Vehicle Routing With Real- Time Traffic Information IEEE Transactions on Intelligent Transportation Systems, Vol. 6, No. 2, June 2005. 25. Lam, T. C., and Small, K.A. The value of time and reliability: measurement from a value pricing experiment. Transportation Research Part E: Logistics and Transportation Review 37(2-3). (2001): 231-251. 26. Map Projections - Part I, Raben Systems, inc. http://www.raben.com/articles/mapprojections/part_1_1.html. (May 29, 2006) 27. Margiotta, R. and R. Taylor. Traffic congestion and reliability: making the connection with operations. Part 1; Measuring and tracking reliability. ITE journal. Vol. 76, No. 2 (Feb. 2006) p. 24-28. ISSN: 0162-8178. 28. Schrank, D. L. and Lomax, T. J. The 2005 urban mobility report. Texas A & M University, Texas Transportation Institute (TTI). College Station: TX, 2005 29. Tarantilis, C.D., Kiranoudis, C.T. Using a spatial decision support system for solving the vehicle routing problem. Information & Management. 39 (2002) 359-375 30. Teply, S. Accuracy of delay surveys at signalized intersections. Transportation Research Record. No. 1225 (1989):24-32 31. Texas State. Texas Department of Transportation (TxDOT). Test Procedures (400-A Series) Tex-440-A, Initial Time of Set of Fresh Concrete. http://www.dot.state.tx.us/services/construction/test_procedures/tms_series.htm?series= 400A. (June 22, 2006). 32. Traffic Engineering Handbook, 3rd edition, Institute of Traffic Engineers (ITE), 1965 33. Transportation Research Board (TRB). Highway capacity manual 2000. Corp Author(s): National Research Council (U.S.).; Transportation Research Board. Washington, D.C., 2000. 34. U. S. Censes Bureau, Ranking Tables for Metropolitan Areas. http://www.census.gov/population/www/cen2000/phc-t3.html. (November 26, 2006) 35. U.S. Geological Survey (USGS). Digital Orthophoto Quadrangles. http://nationalmap.gov/digitalbackyard/doqbkyd.html. (May 29, 2006a) 36. U. S. Geological Survey (USGS). Geographic Information Systems. http://erg.usgs.gov/isb/pubs/gis_poster/index.html. (May 29, 2006b). 119 37. U. S. Department of Transportation (USDOT). Transportation Statistics Annual Report, November 2005. Research and Innovative Technology Administration, Bureau of Transportation Statistics (B. o. T. S), Washington: D. C., 2005 38. U. S. Department of Transportation (USDOT). Deployment statistics, Intelligent Transportation Systems, U. S. Department of Transportation. http://www.itsdeployment.its.dot.gov/overview.asp. (November 26, 2006) 39. What is GPS? Garmin Ltd., http://www.garmin.com/aboutGPS/. (July 26, 2006) 120 APPENDIX A Table A1 compares the route found by GoogleTM map with the time-dependent GIS models developed in this research and the actual routes. Table A2 compares the route found by Map QuestTM with the developed GIS models and the actual routes. In Table A1 and A2 some of the rows are left blank below the GoogleTM and Map QuestTM route vales because these internet tools did not have origin or destination addresses in their GIS maps. This may be a result of newly constructed roads in the COA area that may not have been updated in those internet tools, yet. The Map QuestTM has the capabilities to find shortest time as well as shortest distance routes. Table A1 Comparison of Actual Route, Model Routes, GoogleTM Routes Actual Route Shortest Time Model Shortest Time Model GoogleTM Shortest Time Route Difference in Difference in Difference in Route number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distanc e (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 2.540 5.12 2.540 5.30 0.000 -0.183 2.540 5.30 0.000 -0.18 2.500 6.00 0.040 -0.88 1 Return 2.540 5.02 2.540 5.00 0.000 0.017 2.540 5.00 0.000 0.02 2.500 6.00 0.040 -0.98 Haul 2.540 4.72 2.540 5.38 0.000 -0.667 2.540 5.38 0.000 -0.67 2.500 6.00 0.040 -1.28 2 Return 2.540 4.95 2.540 5.00 0.000 -0.050 2.540 5.00 0.000 -0.05 2.500 6.00 0.040 -1.05 Haul 8.301 16.27 8.301 14.58 0.000 1.683 8.301 14.58 0.000 1.68 - - - - 3 Return 8.301 13.87 8.301 14.15 0.000 -0.283 8.301 14.15 0.000 -0.28 - - - - Haul 4.760 12.98 4.700 11.68 0.060 1.300 4.751 11.02 0.009 1.97 4.500 10.00 0.260 2.98 4 Return 4.760 12.53 4.700 11.42 0.060 1.117 4.751 10.75 0.009 1.78 4.500 10.00 0.260 2.53 Haul 4.760 13.60 4.700 11.82 0.060 1.783 4.751 11.18 0.009 2.42 4.500 10.00 0.260 3.60 5 Return 4.760 12.32 4.700 11.50 0.060 0.817 4.751 10.87 0.009 1.45 4.500 10.00 0.260 2.32 Haul 4.528 10.33 4.518 9.42 0.009 0.917 4.518 9.42 0.009 0.92 4.600 10.00 -0.072 0.33 6 Return 4.528 11.52 4.518 9.15 0.009 2.367 4.518 9.15 0.009 2.37 4.600 10.00 -0.072 1.52 Haul 4.528 13.67 4.518 9.38 0.009 4.283 4.518 9.38 0.009 4.28 4.600 10.00 -0.072 3.67 7 Return 4.528 13.65 4.518 9.12 0.009 4.533 4.518 9.12 0.009 4.53 4.600 10.00 -0.072 3.65 Haul 4.379 11.95 4.379 9.53 0.000 2.417 4.370 9.53 0.009 2.42 4.600 9.00 -0.221 2.95 8 Return 4.379 10.05 4.379 9.38 0.000 0.667 4.370 9.38 0.009 0.67 4.600 9.00 -0.221 1.05 Haul 7.369 9.57 5.800 11.52 1.569 -1.950 7.369 9.55 0.000 0.02 8.200 12.00 -0.831 -2.43 9 Return 7.699 11.23 5.800 11.80 1.899 -0.567 7.699 10.88 0.000 0.35 8.200 12.00 -0.501 -0.77 Haul 3.219 6.42 3.219 5.58 0.000 0.833 3.219 5.58 0.000 0.83 3.400 9.00 -0.181 -2.58 10 Return 3.219 5.87 3.219 5.58 0.000 0.283 3.219 5.58 0.000 0.28 3.400 9.00 -0.181 -3.13 Haul 3.858 8.47 3.588 6.83 0.270 1.633 3.589 6.58 0.269 1.88 3.400 9.00 0.458 -0.53 11 Return 3.858 8.22 3.588 6.83 0.270 1.383 3.589 6.58 0.269 1.63 3.400 9.00 0.458 -0.78 Haul 3.858 7.87 3.588 6.83 0.270 1.033 3.589 6.60 0.269 1.27 3.400 9.00 0.458 -1.13 12 Return 3.858 9.17 3.588 6.83 0.270 2.333 3.589 6.58 0.269 2.58 3.400 9.00 0.458 0.17 12 1 Table A1 Comparison of Actual Route, Model Routes, GoogleTM Routes (cont?d.) Actual Route Shortest Time Model Shortest Time Model GoogleTM Shortest Time Route Difference in Difference in Difference in Route number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 4.265 10.37 4.265 9.78 0.000 0.583 4.265 9.78 0.000 0.58 4.600 9.00 -0.335 1.37 13 Return 4.471 10.82 4.464 10.30 0.007 0.517 4.464 10.30 0.007 0.52 4.400 9.00 0.071 1.82 Haul 4.753 10.87 4.753 10.17 0.000 0.700 4.753 10.17 0.000 0.70 4.800 10.00 -0.047 0.87 14 Return 4.753 9.57 4.753 10.17 0.000 -0.600 4.753 10.17 0.000 -0.60 4.800 10.00 -0.047 -0.43 Haul 4.098 9.40 4.098 8.62 0.000 0.783 4.098 8.62 0.000 0.78 4.400 9.00 -0.302 0.40 15 Return 4.098 7.80 4.098 8.72 0.000 -0.917 4.098 8.72 0.000 -0.92 4.400 9.00 -0.302 -1.20 Haul 4.098 8.72 4.098 8.72 0.000 0.000 4.098 8.62 0.000 0.10 4.400 9.00 -0.302 -0.28 16 Return 4.098 8.83 4.098 8.80 0.000 0.033 4.098 8.80 0.000 0.03 4.400 9.00 -0.302 -0.17 Haul 3.858 7.83 3.588 6.83 0.270 1.000 3.589 6.58 0.269 1.25 3.400 9.00 0.458 -1.17 17 Return 3.858 8.67 3.588 6.83 0.270 1.833 3.589 6.60 0.269 2.07 3.400 9.00 0.458 -0.33 Haul 3.398 6.50 3.398 6.97 0.000 -0.467 3.398 6.97 0.000 -0.47 3.400 8.00 -0.002 -1.50 18 Return 3.398 7.33 3.398 7.02 0.000 0.317 3.398 7.02 0.000 0.32 3.400 8.00 -0.002 -0.67 Haul 3.287 8.25 3.287 6.82 0.000 1.433 3.287 6.82 0.000 1.43 3.300 6.00 -0.013 2.25 19 Return 3.363 6.88 3.363 7.00 0.000 -0.117 3.378 6.68 -0.015 0.20 - - - - Haul 3.990 8.27 3.990 8.45 0.000 -0.183 3.990 8.45 0.000 -0.18 4.000 8.00 -0.010 0.27 20 Return 3.960 10.52 3.960 8.42 0.000 2.100 3.990 8.42 -0.030 2.10 4.000 8.00 -0.040 2.52 Haul 1.877 4.47 1.877 3.55 0.000 0.917 1.877 3.55 0.000 0.92 1.900 5.00 -0.023 -0.53 21 Return 1.877 3.68 1.877 3.78 0.000 -0.100 1.877 3.78 0.000 -0.10 1.900 5.00 -0.023 -1.32 Haul 2.800 5.87 2.800 6.02 0.000 -0.150 2.800 6.02 0.000 -0.15 3.400 6.00 -0.600 -0.13 22 Return 2.800 4.47 2.800 5.02 0.000 -0.550 2.800 5.02 0.000 -0.55 3.400 6.00 -0.600 -1.53 Haul 1.877 4.67 1.877 3.57 0.000 1.100 1.877 3.57 0.000 1.10 1.900 5.00 -0.023 -0.33 23 Return 1.877 3.40 1.877 3.78 0.000 -0.383 1.877 3.78 0.000 -0.38 1.900 5.00 -0.023 -1.60 Haul 6.384 12.08 6.046 12.62 0.338 -0.533 6.384 11.77 0.000 0.32 6.100 14.00 0.284 -1.92 24 Return 6.384 12.63 6.046 12.33 0.338 0.300 6.384 11.70 0.000 0.93 6.100 14.00 0.284 -1.37 12 2 Table A1 Comparison of Actual Route, Model Routes, GoogleTM Routes (cont?d.) Actual Route Shortest Time Model Shortest Time Model GoogleTM Shortest Time Route Difference in Difference in Difference in Route number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 7.677 10.80 7.099 14.72 0.577 -3.917 7.677 10.05 0.000 0.75 - - - - 25 Return 8.111 11.10 7.099 14.60 1.012 -3.500 8.111 10.93 0.000 0.17 - - - - Haul 3.490 6.73 3.490 6.03 0.000 0.700 3.490 6.03 0.000 0.70 3.500 9.00 -0.010 -2.27 26 Return 3.490 5.70 3.490 6.03 0.000 -0.333 3.490 6.03 0.000 -0.33 3.500 9.00 -0.010 -3.30 Haul 7.677 10.37 7.099 14.85 0.577 -4.483 7.677 10.20 0.000 0.17 - - - - 27 Return 8.111 11.73 7.099 15.17 1.012 -3.433 7.677 11.63 0.434 0.10 - - - - Haul 2.800 5.77 2.800 6.02 0.000 -0.250 2.800 6.02 0.000 -0.25 3.400 6.00 -0.600 -0.23 28 Return 2.800 5.40 2.800 5.02 0.000 0.383 2.800 5.02 0.000 0.38 3.400 6.00 -0.600 -0.60 Haul 3.442 6.53 3.442 5.88 0.000 0.650 3.442 5.88 0.000 0.65 3.400 9.00 0.042 -2.47 29 Return 3.442 6.13 3.442 5.88 0.000 0.250 3.442 5.88 0.000 0.25 3.400 9.00 0.042 -2.87 Haul 3.490 7.43 3.490 6.07 0.000 1.367 3.490 6.07 0.000 1.37 3.500 9.00 -0.010 -1.57 30 Return 3.490 7.18 3.490 6.02 0.000 1.167 3.490 6.02 0.000 1.17 3.500 9.00 -0.010 -1.82 Haul 3.858 9.03 3.588 6.82 0.270 2.217 3.589 6.57 0.269 2.47 3.400 9.00 0.458 0.03 31 Return 3.858 7.60 3.588 6.82 0.270 0.783 3.589 6.58 0.269 1.02 3.400 9.00 0.458 -1.40 Haul 7.671 11.93 7.671 11.85 0.000 0.083 7.671 11.85 0.000 0.08 8.600 17.00 -0.929 -5.07 32 Return* - - - - - - - - - - - - - - Haul 4.399 11.63 4.390 9.93 0.009 1.700 4.390 9.93 0.009 1.70 4.400 9.00 -0.001 2.63 33 Return 4.399 10.73 4.390 9.90 0.009 0.833 4.390 9.90 0.009 0.83 4.400 9.00 -0.001 1.73 Haul 7.952 15.20 7.615 14.50 0.338 0.700 7.952 14.07 0.000 1.13 7.800 19.00 0.152 -3.80 34 Return 7.952 15.87 7.615 14.65 0.338 1.217 7.952 14.12 0.000 1.75 7.800 19.00 0.152 -3.13 Haul 4.760 12.70 4.672 11.68 0.088 1.017 4.751 11.02 0.009 1.68 4.600 9.00 0.160 3.70 35 Return* - - - - - - - - - - - - - - Haul 7.952 13.97 7.615 14.65 0.338 -0.683 7.952 14.12 0.000 -0.15 7.800 19.00 0.152 -5.03 36 Return 7.624 17.67 7.615 14.72 0.009 2.950 7.952 14.22 -0.328 3.45 7.800 19.00 -0.176 -1.33 Cumulative Values 317.668 652.43 306.779 619.70 10.889 32.73 315.334 591.18 2.334 61.25 267.700 594.00 -1.572 -22.58 Average Values 4.540 9.32 4.383 8.85 0.156 0.47 4.505 8.45 0.033 0.88 4.249 9.43 -0.025 -0.36 Standard Deviation 1.828 3.31 1.657 3.33 0.346 1.55 1.838 2.88 0.108 1.12 1.606 3.31 0.314 2.08 Note : * = Truck was stationary along route for non-traffic related reasons. 12 3 Table A2 Comparison of Actual Route, Model Routes, Map QuestTM Routes Actual Route Shortest Route Model Shortest Time Model Map QuestTM Shortest Time Route Map QuestTM Shortest Distance Route Difference in Difference in Difference in Difference in Route number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 2.540 5.12 2.540 5.30 0.000 -0.183 2.540 5.30 0.000 -0.18 2.510 6.00 0.030 -0.88 2.510 6.00 0.030 -0.88 1 Return 2.540 5.02 2.540 5.00 0.000 0.017 2.540 5.00 0.000 0.02 2.510 6.00 0.030 -0.98 2.510 6.00 0.030 -0.98 Haul 2.540 4.72 2.540 5.38 0.000 -0.667 2.540 5.38 0.000 -0.67 2.510 6.00 0.030 -1.28 2.510 6.00 0.030 -1.28 2 Return 2.540 4.95 2.540 5.00 0.000 -0.050 2.540 5.00 0.000 -0.05 2.510 6.00 0.030 -1.05 2.510 6.00 0.030 -1.05 Haul 8.301 16.27 8.301 14.58 0.000 1.683 8.301 14.58 0.000 1.68 7.670 16.00 0.631 0.27 7.670 16.00 0.631 0.27 3 Return 8.301 13.87 8.301 14.15 0.000 -0.283 8.301 14.15 0.000 -0.28 7.670 16.00 0.631 -2.13 7.670 16.00 0.631 -2.13 Haul 4.760 12.98 4.700 11.68 0.060 1.300 4.751 11.02 0.009 1.97 4.400 10.00 0.360 2.98 4.400 11.00 0.360 1.98 4 Return 4.760 12.53 4.700 11.42 0.060 1.117 4.751 10.75 0.009 1.78 4.400 10.00 0.360 2.53 4.400 11.00 0.360 1.53 Haul 4.760 13.60 4.700 11.82 0.060 1.783 4.751 11.18 0.009 2.42 4.400 10.00 0.360 3.60 4.400 11.00 0.360 2.60 5 Return 4.760 12.32 4.700 11.50 0.060 0.817 4.751 10.87 0.009 1.45 4.400 10.00 0.360 2.32 4.400 11.00 0.360 1.32 Haul 4.528 10.33 4.518 9.42 0.009 0.917 4.518 9.42 0.009 0.92 4.490 10.00 0.038 0.33 4.490 11.00 0.038 -0.67 6 Return 4.528 11.52 4.518 9.15 0.009 2.367 4.518 9.15 0.009 2.37 4.490 10.00 0.038 1.52 4.490 11.00 0.038 0.52 Haul 4.528 13.67 4.518 9.38 0.009 4.283 4.518 9.38 0.009 4.28 4.490 10.00 0.038 3.67 4.490 11.00 0.038 2.67 7 Return 4.528 13.65 4.518 9.12 0.009 4.533 4.518 9.12 0.009 4.53 4.490 10.00 0.038 3.65 4.490 11.00 0.038 2.65 Haul 4.379 11.95 4.379 9.53 0.000 2.417 4.370 9.53 0.009 2.42 4.430 10.00 -0.051 1.95 4.430 10.00 -0.051 1.95 8 Return 4.379 10.05 4.379 9.38 0.000 0.667 4.370 9.38 0.009 0.67 4.430 10.00 -0.051 0.05 4.430 10.00 -0.051 0.05 Haul 7.369 9.57 5.800 11.52 1.569 -1.950 7.369 9.55 0.000 0.02 8.100 11.00 -0.731 -1.43 5.640 16.00 1.729 -6.43 9 Return 7.699 11.23 5.800 11.80 1.899 -0.567 7.699 10.88 0.000 0.35 8.100 11.00 -0.401 0.23 5.640 16.00 2.059 -4.77 Haul 3.219 6.42 3.219 5.58 0.000 0.833 3.219 5.58 0.000 0.83 3.140 8.00 0.079 -1.58 3.140 9.00 0.079 -2.58 10 Return 3.219 5.87 3.219 5.58 0.000 0.283 3.219 5.58 0.000 0.28 3.140 8.00 0.079 -2.13 3.140 9.00 0.079 -3.13 Haul 3.858 8.47 3.588 6.83 0.270 1.633 3.589 6.58 0.269 1.88 3.140 8.00 0.718 0.47 3.140 9.00 0.718 -0.53 11 Return 3.858 8.22 3.588 6.83 0.270 1.383 3.589 6.58 0.269 1.63 3.140 8.00 0.718 0.22 3.140 9.00 0.718 -0.78 Haul 3.858 7.87 3.588 6.83 0.270 1.033 3.589 6.60 0.269 1.27 3.140 8.00 0.718 -0.13 3.140 9.00 0.718 -1.13 12 Return 3.858 9.17 3.588 6.83 0.270 2.333 3.589 6.58 0.269 2.58 3.140 8.00 0.718 1.17 3.140 9.00 0.718 0.17 12 4 Table A2 Comparison of Actual Route, Model Routes, Map QuestTM Routes (cont?d.) Actual Route Shortest Route Model Shortest Time Model Map QuestTM Shortest Time Route Map QuestTM Shortest Time Route Difference in Difference in Difference in Difference in Route number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 4.265 10.37 4.265 9.78 0.000 0.583 4.265 9.78 0.000 0.58 4.210 9.00 0.055 1.37 4.430 10.00 -0.165 0.37 13 Return 4.471 10.82 4.464 10.30 0.007 0.517 4.464 10.30 0.007 0.52 4.210 9.00 0.261 1.82 4.430 10.00 0.041 0.82 Haul 4.753 10.87 4.753 10.17 0.000 0.700 4.753 10.17 0.000 0.70 4.670 10.00 0.083 0.87 4.670 11.00 0.083 -0.13 14 Return 4.753 9.57 4.753 10.17 0.000 -0.600 4.753 10.17 0.000 -0.60 4.670 10.00 0.083 -0.43 4.670 11.00 0.083 -1.43 Haul 4.098 9.40 4.098 8.62 0.000 0.783 4.098 8.62 0.000 0.78 4.220 9.00 -0.122 0.40 4.060 10.00 0.038 -0.60 15 Return 4.098 7.80 4.098 8.72 0.000 -0.917 4.098 8.72 0.000 -0.92 4.220 9.00 -0.122 -1.20 4.060 10.00 0.038 -2.20 Haul 4.098 8.72 4.098 8.72 0.000 0.000 4.098 8.62 0.000 0.10 4.220 9.00 -0.122 -0.28 4.060 10.00 0.038 -1.28 16 Return 4.098 8.83 4.098 8.80 0.000 0.033 4.098 8.80 0.000 0.03 4.220 9.00 -0.122 -0.17 4.060 10.00 0.038 -1.17 Haul 3.858 7.83 3.588 6.83 0.270 1.000 3.589 6.58 0.269 1.25 3.240 8.00 0.618 -0.17 3.240 9.00 0.618 -1.17 17 Return 3.858 8.67 3.588 6.83 0.270 1.833 3.589 6.60 0.269 2.07 3.240 8.00 0.618 0.67 3.240 9.00 0.618 -0.33 Haul 3.398 6.50 3.398 6.97 0.000 -0.467 3.398 6.97 0.000 -0.47 3.300 8.00 0.098 -1.50 3.300 8.00 0.098 -1.50 18 Return 3.398 7.33 3.398 7.02 0.000 0.317 3.398 7.02 0.000 0.32 3.300 8.00 0.098 -0.67 3.300 8.00 0.098 -0.67 Haul 3.287 8.25 3.287 6.82 0.000 1.433 3.287 6.82 0.000 1.43 - - - - - - - - 19 Return 3.363 6.88 3.363 7.00 0.000 -0.117 3.378 6.68 -0.015 0.20 - - - - - - - - Haul 3.990 8.27 3.990 8.45 0.000 -0.183 3.990 8.45 0.000 -0.18 3.940 8.00 0.050 0.27 3.940 9.00 0.050 -0.73 20 Return 3.960 10.52 3.960 8.42 0.000 2.100 3.990 8.42 -0.030 2.10 3.940 8.00 0.020 2.52 3.940 9.00 0.020 1.52 Haul 1.877 4.47 1.877 3.55 0.000 0.917 1.877 3.55 0.000 0.92 1.870 5.00 0.007 -0.53 1.870 5.00 0.007 -0.53 21 Return 1.877 3.68 1.877 3.78 0.000 -0.100 1.877 3.78 0.000 -0.10 1.870 5.00 0.007 -1.32 1.870 5.00 0.007 -1.32 Haul 2.800 5.87 2.800 6.02 0.000 -0.150 2.800 6.02 0.000 -0.15 3.330 6.00 -0.530 -0.13 2.740 6.00 0.060 -0.13 22 Return 2.800 4.47 2.800 5.02 0.000 -0.550 2.800 5.02 0.000 -0.55 3.330 6.00 -0.530 -1.53 2.740 6.00 0.060 -1.53 Haul 1.877 4.67 1.877 3.57 0.000 1.100 1.877 3.57 0.000 1.10 1.870 5.00 0.007 -0.33 1.870 5.00 0.007 -0.33 23 Return 1.877 3.40 1.877 3.78 0.000 -0.383 1.877 3.78 0.000 -0.38 1.870 5.00 0.007 -1.60 1.870 5.00 0.007 -1.60 Haul 6.384 12.08 6.046 12.62 0.338 -0.533 6.384 11.77 0.000 0.32 - - - - - - - - 24 Return 6.384 12.63 6.046 12.33 0.338 0.300 6.384 11.70 0.000 0.93 - - - - - - - - 12 5 Table A2 Comparison of Actual Route, Model Routes, Map QuestTM Routes (cont?d.) Actual Route Shortest Route Model Shortest Time Model Map QuestTM Shortest Time Route Map QuestTM Shortest Time Route Difference in Difference in Difference in Difference in Route number Haul Type Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Distance (mile) Predicted Travel Time (min) Distance (mile) Travel Time (min) Haul 7.677 10.80 7.099 14.72 0.577 -3.917 7.677 10.05 0.000 0.75 6.850 8.00 0.827 2.80 6.850 9.00 0.827 1.80 25 Return 8.111 11.10 7.099 14.60 1.012 -3.500 8.111 10.93 0.000 0.17 6.850 8.00 1.261 3.10 6.850 9.00 1.261 2.10 Haul 3.490 6.73 3.490 6.03 0.000 0.700 3.490 6.03 0.000 0.70 3.450 10.00 0.040 -3.27 3.450 10.00 0.040 -3.27 26 Return 3.490 5.70 3.490 6.03 0.000 -0.333 3.490 6.03 0.000 -0.33 3.450 10.00 0.040 -4.30 3.450 10.00 0.040 -4.30 Haul 7.677 10.37 7.099 14.85 0.577 -4.483 7.677 10.20 0.000 0.17 6.850 8.00 0.827 2.37 6.850 9.00 0.827 1.37 27 Return 8.111 11.73 7.099 15.17 1.012 -3.433 7.677 11.63 0.434 0.10 6.850 8.00 1.261 3.73 6.850 9.00 1.261 2.73 Haul 2.800 5.77 2.800 6.02 0.000 -0.250 2.800 6.02 0.000 -0.25 3.300 6.00 -0.500 -0.23 2.740 8.00 0.060 -2.23 28 Return 2.800 5.40 2.800 5.02 0.000 0.383 2.800 5.02 0.000 0.38 3.300 6.00 -0.500 -0.60 2.740 8.00 0.060 -2.60 Haul 3.442 6.53 3.442 5.88 0.000 0.650 3.442 5.88 0.000 0.65 3.380 9.00 0.062 -2.47 3.380 9.00 0.062 -2.47 29 Return 3.442 6.13 3.442 5.88 0.000 0.250 3.442 5.88 0.000 0.25 3.380 9.00 0.062 -2.87 3.380 9.00 0.062 -2.87 Haul 3.490 7.43 3.490 6.07 0.000 1.367 3.490 6.07 0.000 1.37 3.450 10.00 0.040 -2.57 3.450 10.00 0.040 -2.57 30 Return 3.490 7.18 3.490 6.02 0.000 1.167 3.490 6.02 0.000 1.17 3.450 10.00 0.040 -2.82 3.450 10.00 0.040 -2.82 Haul 3.858 9.03 3.588 6.82 0.270 2.217 3.589 6.57 0.269 2.47 3.240 8.00 0.618 1.03 3.240 9.00 0.618 0.03 31 Return 3.858 7.60 3.588 6.82 0.270 0.783 3.589 6.58 0.269 1.02 3.240 8.00 0.618 -0.40 3.240 9.00 0.618 -1.40 Haul 7.671 11.93 7.671 11.85 0.000 0.083 7.671 11.85 0.000 0.08 8.620 16.00 -0.949 -4.07 7.520 19.00 0.151 -7.07 32 Return* - - - - - - - - - - - - - - - - - - Haul 4.399 11.63 4.390 9.93 0.009 1.700 4.390 9.93 0.009 1.70 4.370 10.00 0.029 1.63 4.350 11.00 0.049 0.63 33 Return 4.399 10.73 4.390 9.90 0.009 0.833 4.390 9.90 0.009 0.83 4.370 10.00 0.029 0.73 4.350 11.00 0.049 -0.27 Haul 7.952 15.20 7.615 14.50 0.338 0.700 7.952 14.07 0.000 1.13 7.590 19.00 0.362 -3.80 7.900 20.00 0.052 -4.80 34 Return 7.952 15.87 7.615 14.65 0.338 1.217 7.952 14.12 0.000 1.75 7.590 19.00 0.362 -3.13 7.900 20.00 0.052 -4.13 Haul 4.760 12.70 4.672 11.68 0.088 1.017 4.751 11.02 0.009 1.68 4.710 10.00 0.050 2.70 4.500 11.00 0.260 1.70 35 Return* - - - - - - - - - - - - - - - - - - Haul 7.952 13.97 7.615 14.65 0.338 -0.683 7.952 14.12 0.000 -0.15 7.590 19.00 0.362 -5.03 7.900 20.00 0.052 -6.03 36 Return 7.624 17.67 7.615 14.72 0.009 2.950 7.952 14.22 -0.328 3.45 7.590 19.00 0.034 -1.33 7.900 20.00 -0.276 -2.33 Cumulative values 317.668 652.43 306.779 619.70 10.889 32.73 315.334 591.18 2.334 61.25 288.050 620.00 10.202 -7.42 280.520 676.00 17.732 -63.42 Average Values 4.540 9.32 4.383 8.85 0.156 0.47 4.505 8.45 0.033 0.88 4.364 9.39 0.155 -0.11 4.250 10.24 0.269 -0.96 Standard Deviation 1.828 3.31 1.657 3.33 0.346 1.55 1.838 2.88 0.108 1.12 1.753 3.32 0.414 2.09 1.668 3.65 0.434 2.20 Note : * = Truck was stationary along route for non-traffic related reasons. 12 6 127 The GoogleTM map routes provide longer travel distances than the actual route distances while also predicting longer times to travel that distance than the actual routes. The standard deviation value for the actual travel time was same as GoogleTM map routes. The standard deviation of shortest route model travel time is less than GoogleTM map routes. This proves that the GoogleTM map routes have more variation in travel time prediction than the GIS shortest time model. Also from table A2 it is eveident that the Map QuestTM route vales exhibits longer travel time in both ite shortest time and distance routes. The Map QuestTM shortest distance route provides some savings in terms of distance over the actual routes. But the Map QuestTM shorest time routes take more time than the actual route time. The standard deviation value for distance is almost same for actual, GIS models, and Map QuestTM routes. But the standard deviation value for time of Map QuestTM shortest time routes is more than the GIS shortest model route. This proves that the Map QuestTM routes have more variation in travel time prediction than GIS shortest time model.