Development of a Custom Data Acquisition System for the Study of Vehicle Dynamics in Longer Combination Vehicles by Jameson Colbert A thesis submitted to the Graduate Faculty of Auburn University in partial ful llment of the requirements for the Degree of Master of Science Auburn, Alabama August 2, 2014 Keywords: Heavy Truck Vehicle Dynamics, Vehicle Simulation, Data Acquisition Copyright 2014 by Jameson Colbert Approved by David Bevly, Chair, Albert J. Smith Jr. Professor of Mechanical Engineering David Beale, Professor of Mechanical Engineering Song-yul Choe, Associate Professor of Mechanical Engineering Thaddeus Roppel, Associate Professor of Electrical and Computer Engineering Abstract This thesis details the development, deployment, and veri cation of a custom data acquisition system for the purpose of studying vehicle dynamics in a triple trailer Longer Combination Vehicle (LCV). In addition to the data acquisition the thesis details the sim- ulation e orts that were undertaken to both verify the experimental data as well as assess the stability of the vehicle itself. The research was part of an e ort to assess the viability of widening the available roadways that are currently accessible to LCV trailers. This project undertook the task of fully instrumenting a triple trailer LCV with a package of more than 35 sensors and implementation of a custom Data Acquisition System (DAQ) to log over 200 channels coming from the aforementioned sensors. Once out tted with the sensor package, the LCV was put through a variety of dynamic tests including lane changes and constant radius turns in an attempt to capture various dynamic characteristics of the vehicle. A series of simulations were run to match the maneuvers undertaken during the experi- mental phase. That data was then compared to ensure that the simulation did indeed agree with the experimental data. Once in agreement the simulations were expanded to speeds that were not able to be achieved experimentally due to safety concerns. The last element of the simulation was a comparison between the LCV triple and a standard double trailer heavy truck as seen on the highways today. The LCV under test behaved as expected given the prior research into LCV dynamics. Additionally the simulation and the experimental data were shown to agree. The simulation exposed the instability of the vehicle at speeds easily expected on the highways. When comparing the triple LCV to the double it was shown that at lower speeds the rst four units behaved similarly, but as the speeds increased the e ects of the third trailer were shown in the responses of the second. Finally this thesis shows that there is a need for heavy ii precautions before allowing triple LCV to traverse the highway roads. At lower speeds the vehicle is safe but increasing that speed to that of the standard highway speed shows that the vehicle will respond with undesirable outputs. iii Acknowledgments There are many people to thank individually that helped me get to this stage in both my education and my life. The person that is at the top of that list is my wife, Katherine. She has stood by and endured the long hours, missed deadlines, and setbacks that have accompanied this process. Without her, I would not be who I am today. Additionally I would like to thank my father for rst inspiring me to become an engineer, you taught me to never take anything at face value and always strive to understand the ?how? with everything in life. That skill has enabled me to achieve things that I didn?t think I would be able to. Academically, I would like to speci cally thank Dr. David Bevly for advising my through this process of graduate school and having faith in me from the beginning when others did not. Additionally thanks go out to all of my engineering professors, speci cally those in the Mechanical Engineering Department who taught me how to nd the answers to the questions that we face not only in the classroom but in life. Lastly, I would like to thank the members of the GPS and Vehicle Dynamics Laboratory for their collaboration and background knowledge that was provided on not only this project but all projects I worked on during graduate studies. Particular thanks are expressed to Dr. David Broderick, Ryan Hill, and Eric Broshears for assisting in the design, fabrication, and construction of the complex data acquisition system that was implemented during this project. iv Table of Contents Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi 1 Introduction & Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 1.2.1 LCV Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 1.2.2 LCV Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2 Test Vehicle Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 2.1 Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.1.1 Displacement Measurements . . . . . . . . . . . . . . . . . . . . . . . 26 2.1.2 Positioning Measurements . . . . . . . . . . . . . . . . . . . . . . . . 27 2.1.3 Accelerations & Rates Measurements . . . . . . . . . . . . . . . . . . 28 2.2 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 2.2.1 CAN Bus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 2.2.2 Sensor Interfacing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2.2.3 Data Logging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 2.3 Hardware Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 3 Experimental Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.1 Test Track . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.2 Test Vehicle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 3.3 Testing Maneuvers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 v 3.3.1 Constant Radius Turn . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.3.2 Single Lane Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 3.3.3 Double Lane Change . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.4 Data Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.4.1 Data Handling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.4.2 Data Formatting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 3.4.3 RTK Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 3.5 Analysis of Experimental Data . . . . . . . . . . . . . . . . . . . . . . . . . 54 3.5.1 Understeer Characteristics . . . . . . . . . . . . . . . . . . . . . . . 54 3.5.2 Rearward Ampli cation . . . . . . . . . . . . . . . . . . . . . . . . . 61 3.5.3 Roll Behavior of LCV Units During 40 mph Double Lane Change . . 64 3.6 Data Quality Issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 4 Vehicle Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 4.1 Vehicle Con guration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.2 Experimental Maneuver Simulation . . . . . . . . . . . . . . . . . . . . . . . 75 4.2.1 Constant Radius Turn . . . . . . . . . . . . . . . . . . . . . . . . . . 75 4.2.2 Single Lane Change . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 4.2.3 Double Lane Change . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 4.3 Experimental vs. Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.4 Double vs. Triple LCV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 4.4.1 Single Lane Change Comparison . . . . . . . . . . . . . . . . . . . . . 93 4.4.2 Double Lane Change Comparison . . . . . . . . . . . . . . . . . . . . 98 4.4.3 Double vs. Triple Conclusions . . . . . . . . . . . . . . . . . . . . . . 103 4.5 Sources of Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 4.6 Improvements for Future Simulations . . . . . . . . . . . . . . . . . . . . . . 104 5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 5.1 Experimental Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 vi 5.2 Simulation Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 5.3 Final Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 5.4 Future Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 A TruckSima174 Con guration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 A.1 TruckSima174 General Overview . . . . . . . . . . . . . . . . . . . . . . . . . . 113 A.2 Model Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 A.2.1 Vehicle Con guration . . . . . . . . . . . . . . . . . . . . . . . . . . 114 A.2.2 Tractor Con guration . . . . . . . . . . . . . . . . . . . . . . . . . . 117 A.2.3 Trailer 1 Con guration . . . . . . . . . . . . . . . . . . . . . . . . . 121 A.2.4 Trailers 2 & 3 Con guration . . . . . . . . . . . . . . . . . . . . . . 126 A.2.5 Con guring the Converter Dollies . . . . . . . . . . . . . . . . . . . . 126 A.3 Model Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 A.4 Maneuver Paths & Con guration . . . . . . . . . . . . . . . . . . . . . . . . 133 A.4.1 The Constant Radius Maneuver . . . . . . . . . . . . . . . . . . . . . 135 A.4.2 The Single Lane Change Maneuver . . . . . . . . . . . . . . . . . . . 136 A.4.3 The Double Lane Change Maneuver . . . . . . . . . . . . . . . . . . . 139 A.4.4 Future Potential Improvements for TruckSima174 Model . . . . . . . . . 140 vii List of Figures 1.1 LCV Triple Test Vehicle Maneuvering Highway Corner . . . . . . . . . . . . . . 2 1.2 Map of Approved Highway Routes for LCVs [1] . . . . . . . . . . . . . . . . . . 3 1.3 Typical Tractor & Single Semi-Trailer Combination [2] . . . . . . . . . . . . . . 6 1.4 Typical LCV Double Trailer Con guration . . . . . . . . . . . . . . . . . . . . . 7 1.5 Converter Dolly with Pintle Hitch [3] . . . . . . . . . . . . . . . . . . . . . . . . 8 1.6 Possible LCV Con gurations [4] . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 1.7 Rearward Ampli cation Response of Double Tank Trailer [5] . . . . . . . . . . . 13 1.8 Rearward Ampli cation Response of Various Vehicle Combinations [3] . . . . . 14 1.9 Low Speed LCV O -Tracking [4] . . . . . . . . . . . . . . . . . . . . . . . . . . 15 1.10 High Speed LCV O -Tracking [4] . . . . . . . . . . . . . . . . . . . . . . . . . . 16 1.11 Rollover Moment Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 2.1 Sensor Placement on LCV Triple . . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.2 Sketch of Axle De ection Measurements . . . . . . . . . . . . . . . . . . . . . . 27 2.3 Celesco String Potentiometers, SR1A-62 (left) and SP1-25 (right) . . . . . . . . 27 2.4 GPS Receivers Novatel ProPak (left) u-Blox Receiver (right) . . . . . . . . . . . 28 viii 2.5 Illustration of Triple-Trailer Frame of Reference . . . . . . . . . . . . . . . . . . 29 2.6 MemSense Inertial Measurement Unit . . . . . . . . . . . . . . . . . . . . . . . 30 2.7 MemSense Inertial Measurement Unit . . . . . . . . . . . . . . . . . . . . . . . 31 2.8 Oxford RT Units RT2500 (left) and RT3100 (right) . . . . . . . . . . . . . . . . 32 2.9 Placement of Advantech Industrial PC . . . . . . . . . . . . . . . . . . . . . . . 33 2.10 Wiring Diagram Illustrating the Connections Between the Sensors and the DAQ 34 2.11 Diagram of CAN Bus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2.12 Inside of a CAN Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 2.13 CAN Message Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 2.14 Illustration of the MOOS Architecture . . . . . . . . . . . . . . . . . . . . . . . 38 2.15 Screen Shot of Asynchronous Log File . . . . . . . . . . . . . . . . . . . . . . . 39 2.16 String Pots Mounted for Axle De ection . . . . . . . . . . . . . . . . . . . . . . 40 2.17 Example of Toolboxes Mounted Underneath Trailers for Sensor Mounting . . . . 40 2.18 Mounted CAN Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 2.19 String Pots Mounted for Articulation Angle . . . . . . . . . . . . . . . . . . . . 42 2.20 Steer Axle String Pots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 2.21 String Pot Mounted for Steer Angle . . . . . . . . . . . . . . . . . . . . . . . . . 43 2.22 GPS Antenna Mounted on the Trailer Roof . . . . . . . . . . . . . . . . . . . . 44 ix 3.1 Aerial View of NCAT Test Track . . . . . . . . . . . . . . . . . . . . . . . . . . 46 3.2 Single Lane Change Maneuver Diagram . . . . . . . . . . . . . . . . . . . . . . 48 3.3 Double Lane Change Maneuver Diagram . . . . . . . . . . . . . . . . . . . . . . 50 3.4 Zones for separating data recorded on the curves from data recorded on the straights. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 3.5 ISO standard frame-of-reference with Y to the left and Z up. . . . . . . . . . . . 52 3.6 Raw vs. Post Processed Angular Rates. . . . . . . . . . . . . . . . . . . . . . . 52 3.7 Constant Radius Maneuver Path Overlay. . . . . . . . . . . . . . . . . . . . . . 55 3.8 Sections of Constant Radius Maneuver. . . . . . . . . . . . . . . . . . . . . . . . 58 3.9 Sections of Constant Radius Maneuver. . . . . . . . . . . . . . . . . . . . . . . . 58 3.10 Tractor 1 Steering Input . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 3.11 Double Lane Change Steering Input (20 mph) . . . . . . . . . . . . . . . . . . . 63 3.12 Double Lane Change Steering Input (40 mph) . . . . . . . . . . . . . . . . . . . 64 3.13 Steering Axle Body Roll . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3.14 Drive Axle Body Roll . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 3.15 Trailer Body Roll . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 3.16 LCV Body Roll (Un-Filtered) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 3.17 LCV Body Roll (Filtered) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 3.18 Lever arm mounting for string pots . . . . . . . . . . . . . . . . . . . . . . . . . 69 x 4.1 Screen Shot of TruckSima174 Environment . . . . . . . . . . . . . . . . . . . . . . 72 4.2 Screen Shot of TruckSima174 Steering Path Input for Double Lane Change . . . . 73 4.3 Steering Wheel Response to No Steering Input . . . . . . . . . . . . . . . . . . 74 4.4 Yaw Response to No Steering Input . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.5 Screen Shot of Animation for Constant Radius Turn . . . . . . . . . . . . . . . 76 4.6 Path of LCV During Constant Radius Turn . . . . . . . . . . . . . . . . . . . . 76 4.7 Roll of Each Unit for 25 mph Constant Radius Turn . . . . . . . . . . . . . . . 77 4.8 Roll of Each Unit for 40 mph Constant Radius Turn . . . . . . . . . . . . . . . 77 4.9 Unit O -Tracking During 40 mph Constant Radius Turn . . . . . . . . . . . . . 78 4.10 Path of LCV During 25 mph Single Lane Change . . . . . . . . . . . . . . . . . 79 4.11 Low Speed O -Tracking During 25 mph Single Lane Change . . . . . . . . . . . 79 4.12 Path of LCV During 45 mph Single Lane Change . . . . . . . . . . . . . . . . . 80 4.13 High Speed O -Tracking During 45 mph Single Lane Change . . . . . . . . . . 80 4.14 Lateral Acceleration During 25 mph Single Lane Change . . . . . . . . . . . . . 81 4.15 Lateral Acceleration During 45 mph Single Lane Change . . . . . . . . . . . . . 81 4.16 Unit Yaw During 25 mph Single Lane Change . . . . . . . . . . . . . . . . . . . 82 4.17 Unit Yaw During 45 mph Single Lane Change . . . . . . . . . . . . . . . . . . . 82 4.18 Path of LCV During 25 mph Double Lane Change . . . . . . . . . . . . . . . . 83 xi 4.19 Path of LCV During 45 mph Double Lane Change . . . . . . . . . . . . . . . . 83 4.20 Low Speed O -Tracking During 25 mph Double Lane Change (1st Lane Change) 84 4.21 High Speed O -Tracking During 45 mph Double Lane Change (1st Lane Change) 84 4.22 Low Speed O -Tracking During 25 mph Double Lane Change (2nd Lane Change) 85 4.23 High Speed O -Tracking During 45 mph Double Lane Change (2nd Lane Change) 85 4.24 Lateral Acceleration During 25 mph Double Lane Change . . . . . . . . . . . . 86 4.25 Lateral Acceleration During 45 mph Double Lane Change . . . . . . . . . . . . 86 4.26 Lateral Acceleration During 60 mph Double Lane Change . . . . . . . . . . . . 87 4.27 Unit Roll During 60 mph Double Lane Change . . . . . . . . . . . . . . . . . . 88 4.28 Steering Input Comparison for Simulation & Experimental Data . . . . . . . . . 89 4.29 Dolly 1 Articulation Angle Comparison for Simulation & Experimental Data . . 90 4.30 Dolly 2 Articulation Angle Comparison for Simulation & Experimental Data . . 90 4.31 Trailer 1 Roll Comparison for Simulation & Experimental Data . . . . . . . . . 91 4.32 Trailer 3 Roll Comparison for Simulation & Experimental Data . . . . . . . . . 92 4.33 45 mph Single Lane Change Double vs. Triple Yaw Comparison . . . . . . . . . 93 4.34 55 mph Single Lane Change Double vs. Triple Yaw Comparison . . . . . . . . . 94 4.35 65 mph Single Lane Change Double vs. Triple Yaw Comparison . . . . . . . . . 94 4.36 45 mph Single Lane Change Double vs. Triple Roll Comparison . . . . . . . . . 95 xii 4.37 55 mph Single Lane Change Double vs. Triple Roll Comparison . . . . . . . . . 95 4.38 65 mph Single Lane Change Double vs. Triple Roll Comparison . . . . . . . . . 96 4.39 45 mph Single Lane Change Double vs. Triple Lateral Acceleration Comparison 97 4.40 55 mph Single Lane Change Double vs. Triple Lateral Acceleration Comparison 97 4.41 65 mph Single Lane Change Double vs. Triple Lateral Acceleration Comparison 98 4.42 45 mph Double Lane Change Double vs. Triple Yaw Comparison . . . . . . . . 99 4.43 55 mph Double Lane Change Double vs. Triple Yaw Comparison . . . . . . . . 99 4.44 60 mph Double Lane Change Double vs. Triple Yaw Comparison . . . . . . . . 100 4.45 45 mph Double Lane Change Double vs. Triple Roll Comparison . . . . . . . . 100 4.46 55 mph Double Lane Change Double vs. Triple Roll Comparison . . . . . . . . 101 4.47 60 mph Double Lane Change Double vs. Triple Roll Comparison . . . . . . . . 101 4.48 45 mph Double Lane Change Double vs. Triple Lateral Acceleration Comparison 102 4.49 55 mph Double Lane Change Double vs. Triple Lateral Acceleration Comparison 102 4.50 60 mph Double Lane Change Double vs. Triple Lateral Acceleration Comparison 103 5.1 Unit Roll During 55 mph Double Lane Change . . . . . . . . . . . . . . . . . . 107 5.2 Unit Roll During 60 mph Double Lane Change . . . . . . . . . . . . . . . . . . 108 A.1 Screen Shot of Simulation Animation . . . . . . . . . . . . . . . . . . . . . . . . 114 A.2 TruckSim Main Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 xiii A.3 TruckSim Vehicle Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 A.4 TruckSim Custom Second Trailer Con guration . . . . . . . . . . . . . . . . . . 117 A.5 Vehicle Con guration Window . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 A.6 Steer Axle Kinematic Con guration Window . . . . . . . . . . . . . . . . . . . . 120 A.7 Steer Axle Suspension Con guration Window . . . . . . . . . . . . . . . . . . . 121 A.8 Steer Axle Roll Sti ness Con guration Window . . . . . . . . . . . . . . . . . . 122 A.9 Drive Axle Kinematic Con guration Window . . . . . . . . . . . . . . . . . . . 122 A.10 Drive Axle Roll Steer Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 A.11 Trailer Main Con guration Window . . . . . . . . . . . . . . . . . . . . . . . . 123 A.12 Trailer 1 Axle Kinematics Window . . . . . . . . . . . . . . . . . . . . . . . . . 125 A.13 Trailer 1 Axle Leaf Suspension Con guration Window . . . . . . . . . . . . . . 125 A.14 Trailer 1 Roll Sti ness Con guration Window . . . . . . . . . . . . . . . . . . . 126 A.15 Trailer 1 Payload Con guration Window . . . . . . . . . . . . . . . . . . . . . . 127 A.16 Converter Dolly Con guration Window . . . . . . . . . . . . . . . . . . . . . . . 129 A.17 Converter Dolly Sprung Mass Con guration Window . . . . . . . . . . . . . . . 130 A.18 Converter Dolly Kinematics Con guration Window . . . . . . . . . . . . . . . . 131 A.19 Converter Dolly Kinematics Con guration Window . . . . . . . . . . . . . . . . 131 A.20 Converter Dolly Roll Sti ness Con guration Window . . . . . . . . . . . . . . . 132 xiv A.21 Unit Yaw During Open-Loop Maneuver . . . . . . . . . . . . . . . . . . . . . . 133 A.22 Steering Wheel Angle During Open-Loop Maneuver . . . . . . . . . . . . . . . . 134 A.23 Screen Shot of Maneuver Setup Screen . . . . . . . . . . . . . . . . . . . . . . . 134 A.24 Constant Radius Path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136 A.25 3D Representation of Gradual Lane Twist . . . . . . . . . . . . . . . . . . . . . 137 A.26 Gradual Lane Twist . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 A.27 Single Lane Change Driver Input . . . . . . . . . . . . . . . . . . . . . . . . . . 138 A.28 Single Lane Change Cone Positions . . . . . . . . . . . . . . . . . . . . . . . . . 138 A.29 Double Lane Change Driver Input . . . . . . . . . . . . . . . . . . . . . . . . . 139 A.30 Double Lane Change Cone Positions . . . . . . . . . . . . . . . . . . . . . . . . 140 xv List of Tables 1.1 E ciency Comparison Single Trailer vs. LCV Triple . . . . . . . . . . . . . . . 4 1.2 Low Speed O -Tracking Results [3] . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.3 Legend for Table 1.2, Table 1.4, and Figure 1.8 [3] . . . . . . . . . . . . . . . . . 17 1.4 High Speed O -Tracking Results [3] . . . . . . . . . . . . . . . . . . . . . . . . 18 1.5 Resulting Dynamics from Understeer/Oversteer Combinations [6] . . . . . . . . 22 2.1 Summary of Measurements and Sensors . . . . . . . . . . . . . . . . . . . . . . 25 2.2 MemSense Performance Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 30 2.3 CrossBow Nav 440 Performance Statistics . . . . . . . . . . . . . . . . . . . . . 31 2.4 Oxford RT2500/3100 Performance Statistics . . . . . . . . . . . . . . . . . . . . 32 3.1 Understeer Gradient Calculations for LCV Tractor . . . . . . . . . . . . . . . . 59 3.2 Understeer Gradient Calculations for LCV Trailer 1 . . . . . . . . . . . . . . . . 60 3.3 Understeer Gradient Calculations for LCV Trailer 2 . . . . . . . . . . . . . . . . 61 3.4 Understeer Gradient Calculations for LCV Trailer 3 . . . . . . . . . . . . . . . . 61 3.5 Rearward Ampli cation for LCV during 40 mph Double Lane Change . . . . . . 67 4.1 Simulation Plot Key . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 A.1 TruckSim Main Set-Up Screen Options . . . . . . . . . . . . . . . . . . . . . . . 115 A.2 TruckSim Vehicle Con guration Options . . . . . . . . . . . . . . . . . . . . . . 116 A.3 TruckSim Custom Solver Options . . . . . . . . . . . . . . . . . . . . . . . . . . 118 A.4 TruckSim Tractor Set-Up Screen Options . . . . . . . . . . . . . . . . . . . . . . 119 A.5 Trailer Con guration Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 A.6 NCAT Vehicle Gross Weight . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127 A.7 Trailer Linkage Con guration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 A.8 Converter Dolly Con guration Options . . . . . . . . . . . . . . . . . . . . . . . 129 A.9 TruckSim Maneuver Set-Up Screen Options . . . . . . . . . . . . . . . . . . . . 135 xvi Chapter 1 Introduction & Background Highway safety has been and for the foreseeable future will always be a topic of great concern and research. The most prominent reason for that are auto accident fatalities. In 2011 alone, a total of 32,367 people died as a result of a car accident[7]. Most individuals nd themselves weary of driving next to or in the vicinity of large trucks as they fear being involved in an accident with one. of the 32,367 fatalities, 3,373 people where killed in accidents in 2011 involving large trucks. Of those 3,373 people, only 16% (553) of those where occupants of the large truck; the rest being either occupants of another vehicle or pedestrians[7]. So when discussing the safety and stability of large vehicles the people most a ected by these results are the individuals driving in the vicinity the large trucks and not actually the occupants of the large trucks itself. To that e ect there needs to be a good amount of certainty that a large vehicle is stable before it is permitted on highway systems in all states. Currently the only large trucks allowed to operate on all highways in the 50 states are single semi-trailer and smaller double trailer units. Tractor trailers are everywhere, single tractors pulling a single trailer are as common as a family sedan on today?s highways and interstates. To some degree the same can be said for a tractor pulling a pair of ?pup trailers? (28 ft semitrailers); this con guration is referred to as as a double. Both of those combinations are used to transport low-density freight across the nation as both are permitted on the highways in all 50 states. The topic of concern and the focus of this thesis will be the assessment of stability of triple-trailer LCVs, and a comparison of the triple-trailer con guration to the double-trailer con guration using computer simulations. The term ?LCV? is an acronym for Longer Combination Vehicle and is used to describe a combination of two or more trailers with a gross weight in excess of 1 80,000 lb. The most common LCV combinations are two trailers where one is longer than 28 ft or a series of three pup trailers. The test vehicle used in this project is that of a single tractor pulling three pup trailers, it can be seen in Figure 1.1. Figure 1.1: LCV Triple Test Vehicle Maneuvering Highway Corner The research done on this project included both experimental results involving the triple- trailer and computer simulations of the same vehicle. The experimental data was obtained by instrumenting the vehicle shown in Figure 1.1; which is one of the vehicles used at the National Center for Asphalt Technologies (NCAT) for asphalt life cycle testing. The vehicle was then put through a variety of standard dynamic maneuvers during which the vehicle?s response was recorded. The methodology and results of the experimental phase is described in Chapter 3. The simulated data was obtained by running numerous simulations in TrukSim, which is a software package developed by Mechanical Simulation Corporation speci cally for simulating the response of large vehicles. This data was then to be compared with the results of the computer simulations in Chapter 4. The research for this thesis was done in conjunction with a larger e ort to characterize and study LCV dynamics, sponsored by the National Transportation Research Center Incorporated (NTRCI) and was a collaborative e ort including both universities and private companies. The report generated from that e ort can be found in [6]. 2 1.1 Background As mentioned above, both single trailers and combinations of two small trailers are permitted on all highways throughout the nation. The end goal of this research is to assess the stability of the triple-trailer LCV in an e ort to help determine whether or not they should also be allowed throughout this nation?s highway system. Currently only a handful of states permit LCVs to operate on their highways, and each of those states have di erent restrictions in place on what type of LCV can be used. The current highways on which LCVs can operate on in the United States is shown in Figure 1.2; as illustrated the options are limited for the triple trailer LCV of 100? or greater in length combination. This is of note since this is the type of LCV that was used in this study and is the form that has the most push to become legal. Figure 1.2: Map of Approved Highway Routes for LCVs [1] 3 One of the driving forces for this research is the push to expand the network of highways and roads that are accessible to LCV?s. The reasoning for this is to increase the e ciency of the commercial vehicles. For instance, examine the case of a single semitrailer versus a LCV con gured with three pup trailers. The increased e ciency can be seen in Table 1.1; for this example the trip length was 15,000 miles and both trucks are assumed to have the same fuel e ciency of 6.5 mpg and using the average cost of one gallon of diesel being a363.951. This represents a fuel cost of a369,117.70 for this simulated trip. This cost savings is signi cant when you think about this on a national scale and not just a single trip, so it is easy to see why there would be a push to permit the use of LCV on most of the highways in the nation. Table 1.1: E ciency Comparison Single Trailer vs. LCV Triple Vehicle Type Height Width Length # of Trailers ft3 Cost per ft3 Single Trailer 10 8 40 1 3,200 a362.85 LCV Triple 10 8 25 3 6,000 a361.52 Di erences: 2,800 a361.34 1.2 Literature Review There are several di erent classi cations of LCVs, each of them having their own unique advantages and disadvantages. Though the LCV has been around for some time there has been little experimental research performed on this family of vehicles. Most of the experimental research to date has been aimed at characterizing the simple single semitrailer con guration, which makes sense do to the ubiquitous nature of that vehicle. This section will go into what has been done in the past to study the LCV as well as establish a fundamental understanding of what components go into an LCV as well as how they go together to form the vehicle. While there was no research found where a triple LCV was instrumented and then un- derwent experimental testing liken to that described in this thesis there were other e orts put forth in the characterization of Longer Combination Vehicles. One of the most helpful 4 pieces found was ?An Overview of the Dynamic Performance Properties of Long Truck Com- binations? [3]. This paper went through and outlined what was to be expected from vehicles with multiple trailers in combination. it formed a general understanding of the dynamics using simulation and gathering of current literature at the time of publication. This paper served as the foundation for what this thesis expanded upon. When searching for a general understanding of the components found in LCV units ?A Factbook of the Mechanical Prop- erties of the Components for Single-Unit and Articulated Heavy Trucks? [8] by Paul Francher et al described each component and what role it played in the dynamics. In addition to the study of how the LCV?s behave there have been several e orts into exploring how to control them. One such e ort is described in ?Performance Characteristics for Automated Driving of Long Heavy Combinations Evaluated in Motion Simulator? [9]. In that publication the authors describe the e orts of formulating multiple control methods for autonomous driving of LCV?s. They achieve this by studying the inputs of numerous experienced LCV drivers given the same vehicle path to drive. This is the path of the future, commercial vehicles are being explored as one of the leading areas for autonomous driving on the highway systems both stateside and internationally. As mentioned the research done for this thesis was done in conjunction with a larger project put on by the National Transportation Research Center Inc. as such there are several prior e orts put on by the center that all lead up to the project for which this thesis is indebted to. The report for the research project that supported this thesis is found in ?U32: Vehicle Stability and Dynamics Longer Combination Vehicles Final Report? [6]. The aforementioned prior e orts are detailed in [10], [11], and [12]. Those projects performed similar experimental and simulation testing that was undertaken for this research on other heavy articulation vehicles. Those e orts lead directly to current research. 5 1.2.1 LCV Overview The basic components for an LCV are a tractor and an assortment of semitrailers. The formal de nition of a semitrailer is a trailer that cannot stand on it?s own and has to be supported by the vehicle which it follows. The most common occurrence of semitrailers is where only one trailer is being supported by a tractor, this is the typical ?18 Wheeler? or ?semi-truck? illustrated in Figure 1.3. Figure 1.3: Typical Tractor & Single Semi-Trailer Combination [2] The other common occurrence is that of two semi-trailers towed by one tractor. The rst trailer is supported by the tractor exactly like it would be if it were the only semi-trailer attached. The di erence here is the second semi-trailer is supported via a converter dolly, that is also attached to the rst trailer. The converter dolly is what connects a semi-trailer to another semi-trailer or anything other than the tractor for that matter. The name converter dolly comes from the fact that the dolly ?converts? the semi-trailer into a full stand alone trailer that can then be towed via the pintle hitch at the front of the dolly. An illustration of the double trailer con guration can be seen in Figure 1.4 and an illustration of a typical converter dolly is shown in Figure 1.5. 6 Figure 1.4: Typical LCV Double Trailer Con guration 7 Figure 1.5: Converter Dolly with Pintle Hitch [3] As those two combinations are the most prominent throughout the nation they are the only combinations most people think possible. There are however ve other conventional combinations. These, along with the two most popular are pictured in Figure 1.6. As mentioned earlier the triple trailer combination is the focus of this research but much of the same dynamic properties apply to all of these vehicles and are either ampli ed or dampened depending on the con guration. 8 Figure 1.6: Possible LCV Con gurations [4] 1.2.2 LCV Dynamics When dealing with any vehicle dynamics project there are always certain behaviors that are identi ed as crucial. Those behaviors then become the center of attention when 9 out tting the vehicle with a sensor package. With the goal being to capture the behaviors that have been identi ed the selection and implementation of the sensor package can be found in Chapter 2. In the case of the triple trailer LCV there are four key areas of concern. Those are, in no particular order; Rearward Ampli cation, Vehicle O -Tracking, Vehicle Rollover, and Vehicle Understeer/Oversteer propensity. The two that pose the greatest complexities in dealing with LCV stability are Vehicle O -Tracking and Rearward Ampli cation. The rollover phenomena is very similar to that of any other vehicle and as such there are numerous approaches to limit the rollover propensity of a vehicle, or in this case a trailer. The same can be said for Oversteer/Understeer; each unit can be examined as a single vehicle typically would. Rearward Ampli cation is the act of dynamic responses growing in magnitude the fur- ther from the the steer axle you are. This is easily visualized with the cracking of a whip; small rapid inputs results in large outputs at the end of the whip. The formal de nition of rearward ampli cation is the ratio of a unit?s maximum amplitude in a trailing unit to its amplitude in the tractor [13]. The most common calculation when considering rearward ampli cation is the lateral accelerations but in theory, any quantity would work. Vehicle o -tracking is de ned as the lateral deviation between the path of the center of the front axle and the path of the centerline of another part of the vehicle [13]. Essentially it is when a following axle fails to follow the path set by the lead axle, this is not only seen in heavy vehicles but most individuals who drive light trucks experience it on a daily basis. Both rearward ampli cation and o -tracking are directly related to trailer length, and both are large factors is the perceived stability of the LCV. The di cult part is that they are inversely proportionate to trailer length. So the longer the trailer the less rearward ampli cation but the more o -tracking and vis versa. When traveling through urban envi- ronments, shorter trailers are desired as they would give you a smaller turning radius with less o -tracking. However, longer trailers are desired on the highways to reduce rearward ampli cation as o -tracking is of little concern on the highways. 10 Rearward Ampli cation One of the of the most common causes of vehicle crashes is lane departure. In 2011 there were 15,307 fatal roadway departure crashes resulting in 16,948 fatalities, which was 51 percent of the fatal crashes in the United States for that year [14]. When a vehicle begins to depart a lane the most common reaction from the driver is to rapidly correct in an attempt to return to the lane. This results in a sudden, high magnitude change in steer angle. This of course initiates a dynamic maneuver; referred to as an impulse steer response. Additionally another leading cause of vehicle crashes is obstacle avoidance. This maneuver is simulated by either a single or double lane change. The vehicle purposefully leaves the lane it was traveling in and in the case of the double lane change return to the original lane after the obstacle. All three of these maneuvers result in a rapid change in steering input and at high speeds the response of the vehicle can become rather violent. Speed is particularly important when dealing with such maneuvers on LCVs, once the speed is in excess of 50 mph the phenomena of rearward ampli cation plays a large roll in the overall dynamics of the LCV [15]. At such speeds mundane and subtle maneuvers experienced at the tractor can result in a rather large dynamic response towards the end of the vehicle train. This can lead to very dangerous outcomes including rollover and lane departure of subsequent units depending on the lateral acceleration and steering input from the leading vehicle. Rearward Ampli cation In uences The largest factor on rearward ampli cation is the length of the trailer. The longer the trailer, the less susceptible it is to rearward ampli cation. Additionally as the number of trailers increase, so does the rearward ampli cation with the added trailer. For example, a triple trailer LCV with trailers measuring 48 ft in length will experience less rearward ampli cation than a triple with 28 ft trailers. Similarly the triple will have less rearward ampli cation than a double with 28 ft trailers [3]. If the trailers are of the same length then 11 simply the more trailers that are added the more rearward ampli cation can be expected. Additionally the magnitude of rearward ampli cation can be in uenced by the frequency of steering input, location of pintle hitch connections between trailers, and the ratio between lateral sti ness on the vehicle tires to the weight of the vehicle [3]. As mentioned the frequency of the steering input plays a signi cant role in the rearward ampli cation characteristics of a vehicle. This correlation is illustrated in Figure 1.7. This graph was produced using data generated by a double tank trailer simulation. As can be seen at low frequencies the amplitudes of the three units is identical, i.e. there is no rearward ampli cation. As the frequency increases there begins to be some separation between the responses of the three units. As the frequency crosses 1 rad/s (0.16 Hz) the rst two units (tractor & rst trailer) begin to separate themselves from the last two units (dolly & second trailer). It is at this point that the lateral acceleration of the second trailer is greater than that of the tractor. Then, at approx 3 rad/s (0.48 Hz) the maximum separation between the response of the tractor and second trailer is experienced. At this frequency the magnitude of the lateral acceleration experienced by the second trailer is nearly double that of the tractor. 12 Figure 1.7: Rearward Ampli cation Response of Double Tank Trailer [5] Though the gure shows that the lateral acceleration is less that what is generated at lower steering frequencies, the trend as it pertains to rearward ampli cation remains similar in more severe maneuvers and raises great concern as a potential cause for rollover. The danger is that since the driver of the tractor is limited to feeling only the lateral accelerations of the tractor, he/she may nd themselves unaware of a potential rollover about to occur. So it is easy to see the danger in rearward ampli cation since the lateral accelerations experienced by the tractor may be well within its limits while the last trailing vehicle may experience lateral accelerations too great for it to handle, causing rollover. Taking these results and exploring the outcome of several di erent con gurations yields Figure 1.8. This plot looks at the ampli cation gain (the ratio of lateral acceleration between 13 the last and rst vehicle in a LCV) as a function of the same steering input frequency. As previously illustrated, the ampli cation is minimal at low frequencies and begins to separate at approximately 1 rad/s (0.16 Hz) with a maximum separation experienced at approximately 3 rad/s (0.48 Hz). Note that the largest gain is experienced by the triple trailer with 28 ft trailers, which is very similar to the vehicle used in testing for this research. The description of each con guration in the study can be found in Table 1.3. Figure 1.8: Rearward Ampli cation Response of Various Vehicle Combinations [3] Rearward Ampli cation Outcomes The most severe result of rearward ampli cation is vehicle rollover. This is due to the increased lateral acceleration in each successive unit which at some point will exceed the vehicle?s static rollover threshold (SRT). The SRT is de ned as the maximum angle which a vehicle can withstand before it or any of the successive units experience rollover. As with any vehicle though, the rollover threshold is e ected by a variety of parameters including 14 but not limited to CG height, suspension components, tire sti ness, track width, and load distribution. Vehicle O -Tracking During low-speed turns, the trailing axle follows a path inside the one set by the lead axle. The lateral o set from the front axle path is proportionate to the wheelbase of the vehicle. This same trait is exhibited in LCVs in that the axles of each unit of the combination track to the inside of the path followed by the preceding axle. This phenomena is what leads to drivers of semi-trucks to have to ?square the corner? when making turns in low speed urban environments. An illustration of low speed o -tracking is shown in Figure 1.9. As previously stated as speed is increased the trailing axle tracks closer to the centerline of the lead axle and will eventually, at high enough speeds track to the outside of the lead axle path. This behavior is illustrated in Figure 1.10. Figure 1.9: Low Speed LCV O -Tracking [4] 15 Figure 1.10: High Speed LCV O -Tracking [4] The level of o -tracking is e ected by the length between each axle and the number of articulation points between the units of the vehicle also in uences o -tracking [3]. One of the key areas of concern is that of on ramps and exit ramps of highways and interstates. With enough o -tracking the trailing vehicle can end up traveling o the inside edge of the ramp [3]. In order to better characterize o -tracking a simulation was run with seven di erent con gurations at low speeds around an exit ramp with a constant radius of 300 ft. Those results are shown in Table 1.2. The abbreviations for the con gurations are detailed in Table 1.3. The ?Swept Path? is the width needed to pass the entire vehicle, assuming a common width of 102 inches [3]. As discussed above, if a vehicle increases speed through the turn the amount of o -tracking diminishes to a point where the trailers will actually begin to track outside of the path set by the lead vehicle. The same simulation was run as above but with a radius of 600 ft and at a speed of 55 mph, those results are shown in Table 1.4. Again, refer to Table 1.3 for abbreviation explanations. In the high speed simulation, all of 16 Table 1.2: Low Speed O -Tracking Results [3] Maximum O -Tracking Swept Path m ft m ft 1 Double-28 0.61 3.20 2.0 10.5 2 Triple-28 0.88 3.47 2.9 11.4 3 Tr/Semi 48 0.98 3.57 3.2 11.7 4 RMD-45/28 1.04 3.63 3.4 11.9 5 RMD-48/28 1.16 3.75 3.8 12.3 6 TPD-45/45 1.49 4.08 4.9 13.4 7 TPD-48/48 1.71 4.30 5.6 14.1 Table 1.3: Legend for Table 1.2, Table 1.4, and Figure 1.8 [3] Semitrailer Semitrailer Short Name Full Name Lengths (ft) Lengths (m) Tr/Semi-48 Tractor with 48 ft Semitrailer 48.0 - 14.6 - Double-28 STAA Double 28.0 28.0 8.5 8.5 RMD 48/28 Rocky Mountain Double 48.0 28.0 14.6 14.6 RMD45/28 Rocky Mountain Double 45.0 28.0 13.7 13.7 TPD 48/48 Turnpike Double 48.0 48.0 14.6 14.6 TPD 45/45 Turnpike Double 45.0 45.0 13.7 13.7 Triple-28 Triple 28.0 28.0 8.5 8.5 the con gurations saw a decrease in o -tracking. Additionally all of the LCV con gurations saw more o -tracking than the typical trailer/semitrailer combination. While interesting, there shouldn?t be much weight put on this as the potential for the high speed o -tracking phenomenon to cause a collision is considered minimal [3]. 17 Table 1.4: High Speed O -Tracking Results [3] O -Tracking m ft 1 Tr/Semi 48 0.16 0.52 2 TPD-48/48 0.34 1.10 3 TPD-45/45 0.38 1.25 4 RMD-48/28 0.41 1.33 5 Double-28 0.44 1.43 6 RMD-45/28 0.44 1.45 7 Triple-28 0.65 2.13 Vehicle Rollover When a vehicle undergoes a dynamic maneuver, weight is transferred from the inside wheel to the outside wheel. If the vehicle has a suspension system then the phenomena of roll is experienced. The aforementioned weight transfer creates a moment about the roll center of the vehicle that is referred to as the roll moment. To counteract this moment, force is transferred from the inner wheels to the outer wheels. This is illustrated in Figure 1.11. Summing the moment about the center of the outer tire contact patch yields Equation (1.1). This is using the small angle approximation of sin? = ?;cos? = 1. 18 Figure 1.11: Rollover Moment Diagram Mayhcg M?hcg +Fzit Mgt2 = 0 (1.1) where Fzi = Inner Vertical Tire Force t = Track Width M = Vehicle Mass ay = Lateral Acceleration hcg = Vertical Height of Center of Gravity (CG) ? = Transverse Slope It is important to note that there were some assumptions made when deriving Equa- tion (1.1), those being that you can neglect the e ect of the vehicle frame and suspension compliance. These neglections are made because rigid chassis behavior is much easier to 19 calculate as you do not need more precise information about the vehicle i.e. suspension sti ness and roll center. As long as the vehicle stays within its rollover threshold, the e ect of this assumption is minimal [6]. The phenomena of roll happens in several stages. First the vertical load on the vehicle is transferred from the inner tires to the outer tires and continues until the vertical load on the inner tires is equal to zero. This point is refereed to as the stability limit of the vehicle. The vehicle will not rollover until one of two things happen. One, the location of the center of gravity (CG) passes outside of the outer tires. Or the stabilizing moment described above is less than the roll inertia generated by the maneuver. This rarely happens in passenger vehicle as they experience what is called yaw divergence typically long before rollover. Yaw divergence is when the tires cannot provide enough lateral force to maintain the maneuver and the tires begin to slide on the road. Vehicle Understeer/Oversteer One of the most common metrics to measure a vehicle?s steering characteristic is by calculating the understeer gradient. As de ned in Equation (1.2), the understeer gradient is the ratio of the deviation from the Ackerman steering angle to the lateral acceleration [16]. K = L R 57:3 ay (1.2) where = Road Wheel Angle (deg) L = Tractor Wheel Base Length (steer axle to drive axle center-line) R = Turn Radius v2 Rg = Lateral Acceleration ay K = Understeer Gradient (deg/g) The factor 57.3 converts from radians to degrees 20 However, when you look to analyze a heavy vehicle with many axles it become much more complicated; but can be expressed as an equivalent two-axle vehicle [17]. The yaw stability of a vehicle is dependent on whether the understeer gradient, K for that vehicle is greater than zero, less than zero, or exactly zero. In the case where K is less than zero the vehicle is said to be understeer. Whereas a positive K characterizes the vehicle as oversteer; if K is zero then the vehicle is said to exhibit neutral steer characteristics [16]. Passenger cars in today?s market avoid making a vehicle that is oversteer as that is a more dangerous case for the everyday driver. In the oversteer case it would take less steer angle to perform a constant radius turn if the driver were to increase speed. It is obvious to see how that would not be the desired condition for even the most experienced drivers to deal with on a daily basis. When dealing with LCV?s, Equation (1.2) can be applied to each unit of the combination in order to determine each unit?s understeer gradient. This is achieved by using the road wheel angle of the steer axle to calculate K for the tractor. The articulation angle between the subsequent trailers and the dolly or tractor to which they are connected is used to calculate K for the other vehicles in the combination [5]. When studying LCV dynamics it is important to remember that since each unit has an independent value for K there can be units that are oversteer at the same time as there are other ones that are understeer and/or neutral steer. Furthermore, as each unit in the combination is unique in that they are loaded and constructed di erently; the e ects of vehicle speed and lateral acceleration on the unit?s understeer gradient are going to be di erent [18, 19]. Being that there are instances where there will be a mixture of understeer and oversteer units in a LCV, it is important to understand what each combination will mean for the dynamics of the vehicle. Since it is very rare to see any vehicle exhibit neutral steer we will only consider the resulting four possible combinations of understeer and oversteer. The dynamics of each combination is described in Table 1.5. 21 Table 1.5: Resulting Dynamics from Understeer/Oversteer Combinations [6] Leading Vehicle Oversteer Understeer Trailing Vehicle Understeer As speed increases toward the critical speed, the articulation gain approaches in nity. This re- sults in a jackknife. System is un- stable at high speed. The vehicle is stable. As speed increases, the articulation angle gain (increase in articulation rel- ative to steering input increase) will approach the ratio of the two units? understeer gradients. Ov ersteer Response depends on whether the ratio of the understeer gradients is greater or less than the ratio of the wheelbases, and the articu- lation gain will go to negative or positive in nity. This results in a jackknife or a swing out, though the di erence will be hard to tell from the driving perspective. The vehicle is unstable at high speed. The articulation gain is initially positive but becomes negative and the trailer swings out. This is an unstable arrangement at high speed. However, at low speeds, the articulation gain is positive, making the vehicle drivable. 22 Chapter 2 Test Vehicle Instrumentation The aim of this chapter is to detail the process of developing and implementing the data acquisition system (DAQ) for use on the LCV triple test vehicle. The goal of the DAQ was to capture those responses outlined earlier in section 1.2.2 at high data rates and store the data in a manner conducive to analysis. Each section will discuss the process of selecting the sensor(s) as well as integrating the sensor into the DAQ. The resulting data les containing the dynamic measurements will facilitate the analysis and comparison of the response of the actual vehicle to that of the simulated vehicle. Through a collaborative e ort with other research partners the following list of responses were chosen to be targeted for measuring: 1. Roll Angle per Lateral Acceleration 2. Yaw Rate per Lateral Acceleration of Tractor and Trailers 3. Path Deviation Error per Trailer Referenced to the Tractor 4. Lateral Acceleration per Steer Angle 5. Yaw Rate per Steer Angle 6. Lateral Acceleration per Trailer 7. Lateral Acceleration of the Tractor 8. Yaw Rate per Trailer 9. Yaw Rate of the Tractor 10. Understeer Gradient of the Tractor and Trailers 23 11. Trailer Motion Damping Time and Distance The above characteristics can be separated into three measurement groups 1. Position Measurements 2. Inertial Measurements 3. Displacement Measurements 2.1 Sensors Sensor selection was a critical part of this project. With such a large vehicle it required a large amount of sensors; luckily this research involved several other research team members outside of Auburn University that were able to help provide the additional sensors needed to instrument the triple. This process included many discussions on not only what needed to be captured but how to capture what was identi ed as the critical responses. In the end a total of 37 individual sensors were installed on the triple-trailer. These sensors produced a total of 206 individual channels of data that were to be recorded by the DAQ. That process is outlined below in Section 2.2. The measurements that were recorded are shown below in tabular form in Table 2.1 as well as in picture form in Figures 2.1. Note that in Table 2.1 a sensor bracketed by ?()? means that the measurement was attempted but for some reason not captured. The channels that are key to characterizing vehicle dynamics were to be recorded at a nominal sampling rate of 100 Hz. Supplementary channels were sampled at lower rates, as indicated in the table. Section 2.2 explains how each sensor?s data was transmitted to a central computer and every single data point was given a time stamp. While the sampling rate was not exactly 100 Hz and not all channels were sampled simultaneously, an experiment prior to data collection showed that the method reliably provided data at the time indicated in each time stamp. 24 Table 2.1: Summary of Measurements and Sensors Unit Quantity Vehicle CAN String Pot GPS IMU GPS & IMU Tractor Horizontal and Vertical Dis- placement of the Steer Axle SP1-25 Horizontal and Vertical Dis- placement of the Second Drive Axle SP1-25 Steer Input at the Steer Col- umn SP1-25 Steer Input, Steer Kingpin SP1-25 Engine Parameters (X) Three-Axis Accelerations and Angular Rates MemSense Position Novatel Trailer 1 Horizontal and Vertical Dis-placement of the Axle SP1-25 Three-Axis Accelerations and Angular Rates CrossBow Position Novatel Trailer 2 Articulation Angle to Lead- ing Converter Dolly SR1A- 62 Horizontal and Vertical Dis- placement of the Axle SP1-25 Three-Axis Accelerations and Angular Rates (RT2500) Position Novatel Trailer 3 Articulation Angle to Lead- ing Converter Dolly SR1A- 62 Horizontal and Vertical Dis- placement of the Axle SP1-25 Three-Axis Accelerations and Angular Rates (RT3200) Position Novatel Each unit of the vehicle had two GPS receivers one for the measurement and one for redundancy. Additionally each unit had an IMU to record accelerations and angular rates, and string potentiometers to measure displacements within the vehicle. The two GPS and 25 Figure 2.1: Sensor Placement on LCV Triple IMUs were separate sensors on the Tractor and Trailer 1; whereas the GPS units and the IMU were integrated on Trailers 2 and 3. 2.1.1 Displacement Measurements String potentiometers or ?string pots? were used to measure dolly-to-trailer articulation angles, lateral and vertical displacement of axles, and steering input. Several of the desired characteristics have to do with vehicle roll and understeer tendencies. In order to fully capture those characteristics it is necessary to understand what the axle of the vehicle is doing at the time of the maneuver. To achieve this, the decision was made to use string pots to measure the linear displacement of each axle in three dimensions. A diagram to show how this was to be achieved is shown below in Figure 2.2. Combining the string pot measurements with the IMUs and GPS measurements the desired characteristic values can be obtained. A total of 27 string pots were used along the vehicle; four were model SR1A- 62 shown to the left in Figure 2.3 used to measure the dolly-to-trailer articulation. These weatherproof units have a stroke length of 62 inches. All other string pots on the vehicle were model SP1-25, shown in Figure 2.3 to the right, with a stroke length of 25 inches. They were installed to measure axle displacement on the test vehicle. The steering input was also measured with a SP1-25 string pot. 26 Figure 2.2: Sketch of Axle De ection Measurements Figure 2.3: Celesco String Potentiometers, SR1A-62 (left) and SP1-25 (right) 2.1.2 Positioning Measurements One of the most crucial measurements for this project is position. If the position mea- surement is not accurate then the entire analysis can be e ected greatly. GPS receivers range 27 wildly in performance as well as cost. The most costly units being the ones used in missile guidance and other various high precision environments. The cheapest being the ones placed in consumer electronics such as cell phones. In an attempt to minimize the error all four of the units were out tted with a Novatel ProPak GPS receiver that provides acceptable performance for such a project. In addition the the Novatel receivers, the tractor and the rst trailer had a u-Blox GPS receiver for redundancy. The second and third trailer used the OxfordRT units for redundancy. A picture of the Novatel receiver as well as the u-Blox receiver can be seen in Figure 2.4, an image of the OxfordRT units is shown below in Section 2.1.3. Figure 2.4: GPS Receivers Novatel ProPak (left) u-Blox Receiver (right) 2.1.3 Accelerations & Rates Measurements Inertial measurements use accelerometers and gyroscopes to measure the rate of change bot laterally and rotationally in anywhere from one to six axes. Inertial Measurement Units (IMUs) contain three accelerometers and three gyroscopes internally aligned in such a fashion to measure both the acceleration and rotational rate along three axes. The IMU is referred to as a six axes since there are two measurements on each axes. When placed near the center of gravity of a vehicle, the output from the IMU can be translated into the yaw, pitch, and 28 roll of the vehicle. Figure 2.5 shows the three axes that are measured by the IMU and which axes translates to yaw, pitch, and roll. Figure 2.5: Illustration of Triple-Trailer Frame of Reference IMU grades range from tactical, which is used in missiles and other very high precision environments all the way to hobby level where the sensors are very cheap and precision is not of the highest priority. Several quality IMUs were borrowed for this project from various sources. An IMU was mounted on each of the units in the combination. A MemSense IMU was obtained to be mounted on the tractor. Being much lower in price than the other sensors, the MemSense o ers exceptional quality for its price. The MemSense is shown in Figure 2.6 along with its technical speci cations in Table 2.2. 29 Table 2.2: MemSense Performance Statistics Figure 2.6: MemSense Inertial Measurement Unit The Crossbow NAV 440, shown in Figure 2.7 is a higher-grade IMU. Table 2.3 shows its technical speci cations. 30 Table 2.3: CrossBow Nav 440 Performance Statistics Figure 2.7: MemSense Inertial Measurement Unit 31 Table 2.4: Oxford RT2500/3100 Performance Statistics The Oxford RT-2500 Inertial and GPS Navigation System, shown in Figure 2.8, to the left includes three angular rate sensors (gyroscopes), three servo-grade accelerometers, the GPS receiver and all the required processing in one very compact box. The internal pro- cessing includes the strap-down algorithms (using a WGS-84 earth model), Kalman ltering and in- ight alignment algorithms. The internal Pentium-class processor runs the QNX real- time operating system to ensure that the outputs are always delivered on time the technical speci cations can be seen in Table 2.4. The Oxford RT-3100, shown in Figure 2.8, to the right is similar to the Oxford RT-2500 model; but promises a little better accuracy in its measurements. Its technical speci cations can be seen in Table 2.4 as well. Figure 2.8: Oxford RT Units RT2500 (left) and RT3100 (right) 32 2.2 Data Acquisition The test vehicle was equipped with numerous sensors that needed to be interfaced with various communication methods. To facilitate this complex data collection with the maximum control over the process, a custom data collection system was developed. The system consisted of an industrial computer with Ubuntu Linux operating system and a wireless (WiFi) router. The Industrial computer ran without a monitor, keyboard, or mouse and connected to the Oxford RT units via the router on an Ethernet network connection. The wireless router allows the data collection process to be monitored and controlled via a laptop computer in the truck cab. It should be noted that no data was sent over the wireless link it was only to monitor and control the data acquisition. All data was recorded directly to the industrial PC and the data acquisition process, and once started was not interrupted by a loss of communication from the laptop in the cab. The location of the PC can be seen in Figure 2.9. Figure 2.9: Placement of Advantech Industrial PC The Advantech PC was wired to each of the sensors either directly or through an on- board network i.e. CAN bus. The interfaces of the Advantech are detailed as follows. a136 Direct UDP/IP communication with the OxfordRT units over CAT6 Ethernet 33 a136 Four Serial (RS-232) ports: { Three Novatel GPS Units for Tractor and Trailer 1&2 { Trailer 1 IMU (Crossbow 440) a136 Six USB Connections { MemSense IMU { RS-232 to USB Converter for Novatel GPS Receiver on Tractor { Two u-Blox GPS Receivers { Two USB-CAN Converters for the Two CAN Buses The wiring between the sensors and the computer is shown in Figure 2.10. Figure 2.10: Wiring Diagram Illustrating the Connections Between the Sensors and the DAQ 2.2.1 CAN Bus Most of the sensors in this project had internal signal conditioning and output a digital signal. Part of this research required the need to build signal conditioning, including an analog-to-digital (A/D) converter, for the string pots. The string pots needed to be connected to an A/D converter, which would convert the signal to digital and then send the message over a Controller Area Network (CAN) network similar to that which is found on most 34 production vehicles. The main driving force behind this decision as opposed to purchasing and using a commercial DAQ with A/D converters attached was controllability. Since the entire system was to be designed and implemented in house, all of the inner workings will be known making any possible troubleshooting easily accomplished. A diagram of the system is in Figure 2.11. Figure 2.11: Diagram of CAN Bus As mentioned, the CAN network is the same network that is standard on vehicles today in order for microcontrollers and devices to communicate with each other within a vehicle without a host computer. The LCV contained two separate CAN buses, one that is already there from the factory and one that was created for the string pots. The factory installed CAN network is in accordance with the SAE J1939 standard. Using the SAE J1939 message list, the vehicle codes were translated and the values were able to be read by the DAQ and subsequently logged. The second network created was for use with the string pots and their associated A/D converters. This path was chosen for multiple reasons; the main two being complete control over design as well as lowering line noise on the string pots by placing the A/D converter close to the string pots themselves. The idea of having complete control over the network is the most critical part of the whole process. This allowed for the creation of a proprietary message set that could not interfere with any other CAN network. An added bene t is that it provided extensive knowledge of the system so that in the event something went wrong 35 it took less time to diagnose and resolve the issue. There were a total of six CAN boxes created, each containing the A/D converter, CAN chip, and circuit board. One of these boxes can be seen in Figure 2.12. Figure 2.12: Inside of a CAN Box Each string pot was connected to the CAN box by a custom DB25 connector that was wired to provide power and ground to each string pot as well as a line for the message. In order to collect the measurements from the string pots, each A/D had to be polled over the network. This was achieved by setting up one of the CAN boxes to send out a message on the CAN bus at 100 Hz that would trigger each of the A/D converters to poll all of the connected string pots. The message set used for the string pots is relatively straightforward; each A/D will send out a status message every second to alert the computer if something is not working. The message structure for the String Pot CAN-bus can be seen in Figure 2.13. 36 Figure 2.13: CAN Message Structure 2.2.2 Sensor Interfacing Code had to be written to receive and interpret the data from each sensor. Some of the code already existed and was used again for this research. The new sensors that had to be interfaced were the MemSense IMU, the u-blox GPS units, and the Oxford RT units. The code for all of the sensor interfaces was written in C++ and ran on the embedded Linux computer on the tractor. The middleware used to communicate all of the sensor data between processes is called Mission Oriented Operating Suite (MOOS) [20]. This middleware uses client-server architecture with a centralized database and provides basic tools for Inter- Process Communication via sockets. This allows for viewing the raw data in real time, and logging the data in a standard format. Each sensor had a corresponding process running that interfaced with that sensor, formated its data, and published it as separate channels to the MOOS database. The MOOS database holds the latest value from each channel on 37 each sensor and allows other processes to access this data. An illustration of this interface process can be seen in Figure 2.14. Figure 2.14: Illustration of the MOOS Architecture 2.2.3 Data Logging Using the laptop in the cab of the truck, the logging was started and stopped via Secure Shell (SSH). Additionally, the data was monitored using MOOSs uMS utility which shows the latest values for each of the sensors in a table. The MOOS utility pLogger connects to the MOOS database and records all changes to all channels to an asynchronous log format. This preserves the data in its raw form with its original timestamps taken by the sensor interface process. This is more di cult to process than synchronous data logs, but provides more accurate timing information. An example of the le format can be seen in Figure 2.15. 38 Figure 2.15: Screen Shot of Asynchronous Log File 2.3 Hardware Installation With the three trailers being nearly identical, the mounting boxes and brackets required were fairly similar. Several sensors needed to be mounted in the tractor?s engine compart- ment. This included one string pot set up on the steer axle, along with two string pots for vertical and horizontal displacement of the kingpin, and lastly a string pot for the angular rotation of the steering shaft. Behind the tractor cab a toolbox with the main CPU was mounted. This box is where the two CAN buses were connected to the industrial computer for data acquisition and processing. Figure 2.9 shows the actual toolbox that was on the tractor along with the location. This box contained the industrial computer, which handled all of the data acquisition as well as various networking devices used for communication throughout the vehicle. As previously stated, string pots were used to measure the vertical and horizontal dis- placement of the axles. One pair of string pots were mounted directly above the axle and 39 another mounted directly horizontal to the axle. This was done to maximize the displace- ments and maximize signal to noise. Figure 2.16 shows these brackets mounted on the LCV with the string pots attached. Note the cable for the string pots is highlighted in red for all of the following pictures. Figure 2.16: String Pots Mounted for Axle De ection Similar to the toolbox mounted directly behind the cab of the tractor. Another toolbox was mounted under each of the trailers. Figure 2.17 shows how they were mounted; a close-up shows the power and serial cables going into and coming out of the box. Figure 2.17: Example of Toolboxes Mounted Underneath Trailers for Sensor Mounting As described above, each of the string pots were connected to a CAN-box that contained the A/D converter. These A/D converters could handle six string pot inputs, and they output 40 their data into the CAN bus that run alongside the trailer combination. Figure 2.18 below shows one of these A/D converter boxes mounted under a trailer. Figure 2.18: Mounted CAN Box To calculate the articulation angle, the longer string pots were mounted on the dolly and attached to the successive trailer. They were crisscrossed in an attempt to get as linear of a measurement as possible and to maximize the displacement. Figure 2.19 shows the actual mounting on the LCV. 41 Figure 2.19: String Pots Mounted for Articulation Angle The tractor?s engine compartment was not only less conducive to taking measurements, but it also required more of them. The displacement of the kingpin on both sides required two more string pots to be mounted on the axle. Additionally, measuring the angular displacement of the steering shaft required a string pot be mounted perpendicular to the shaft. The axle de ection string pots can be seen in Figure 2.20 and the steering shaft in Figure 2.21. Figure 2.20: Steer Axle String Pots 42 Figure 2.21: String Pot Mounted for Steer Angle Each of the GPS receivers requires antennas to be mounted on top of its respective trailer. Shown in Figure 2.22 are the patch antenna on the center of the trailer roof for the u-blox, and the larger antenna on the side of the trailer roof for the Novatel. 43 Figure 2.22: GPS Antenna Mounted on the Trailer Roof 44 Chapter 3 Experimental Results The scope of this chapter is to outline and describe the experimental testing undertaken with the LCV triple at the NCAT test track. The goal of the test plan was to expose the test vehicle to various maneuvers that are designed to incite certain dynamic responses. The dynamic responses of interest were outlined earlier in Section 1.2.2. The responses to these maneuvers were to be captured by use of a custom designed data acquisition system (DAQ) that was detailed in Chapter 2. 3.1 Test Track All of the experimental testing was conducted at NCAT in Opelika, AL. This facility is used for advanced asphalt wear testing and uses a eet of tractor trailer combinations to wear sections of asphalt at accelerated rates. The track located at the facility is one of a kind and o ers a very unique platform for both heavy truck research as well as other automotive research goals that could not be conducted on normal roads and highways. The track is a 1.7 mile long oval comprised of 46 di erent 200-ft. test sections. The two curves the track has an 8a176 banking and use a spiral-curve-spiral con guration. An aerial view of the track can be seen in Figure 3.1. 3.2 Test Vehicle As mentioned, NCAT operates a eet of tractor trailer combinations, the majority of which consist of a lead tractor pulling three sleds loaded with steel plates. However, one of the vehicles was perfectly suited for this research project. It consists of a Freightliner tractor pulling three standard shipping containers; which are loaded down with concrete barriers. 45 Figure 3.1: Aerial View of NCAT Test Track The containers are loaded to greater than the legal limit in order to increase the wear rate of the asphalt. Each container measures 20 ft in length and is riding on a Cheetah container chassis. The rst trailer is obviously connected to the tractor with the second and third trailers being connected via identical silver eagle converter dollies. 3.3 Testing Maneuvers As mentioned above, the vehicle underwent several di erent maneuvers as a part of this testing. This section will describe in detail each of those maneuvers as well as what the maneuver is aimed at achieving with regards to the LCV triple. 3.3.1 Constant Radius Turn Object of Maneuver The object of the constant radius maneuver is to asses several of the vehicle?s steady- state handling characteristics. The main two being the understeer gradient of each trailer as well as the roll steer of the converter dollies. The maneuver was to be repeated at various speeds ranging from 20 to 45 mph. The driver was to have nal say on whether or not each 46 attempt was to be completed. Each speed was to be repeated 10 times to ensure that not only adequate but complete data was recorded. In order for this maneuver to be successful in characterizing the vehicle characteristics the conditions of the maneuver needed to meet most if not all of the following. a136 Consistent Road Wheel Steer Angle a136 Consistent Vehicle Speed a136 Consistent Lateral Acceleration for runs of the same speed Maneuver Path The maneuver is to be performed exactly as it sounds. The vehicle is to travel in a constant radius turn while attempting as best as possible to maintain the conditions listed above. As mentioned earlier in Section 3.1 the curves at the NCAT test track are not true curves except for in the middle of the curve. For this reason, the GPS position data that is recorded during the maneuver will serve to both establish the correct course that the maneuver is performed on but also will de ne the radius that the vehicle travels. 3.3.2 Single Lane Change Object of Maneuver The goal of this test is to capture the response of the vehicle due to obstacle avoidance style maneuver. The target of the test is to show the transient response of the vehicle due to the step change in position. The output from this test will allow for the study of path deviation for all of the units in the vehicle train as well as the settling distances of the subsequent units due to steering inputs at the tractor. Similarly it will allow for the study of rearward ampli cation of the units. The critical parameters for the test are each unit?s lateral acceleration and relative position of each unit to the original path of the tractor. 47 Maneuver Path The path that the vehicle is to follow is exactly like the name of the maneuver suggests. The goal is to begin the test with the vehicle in a steady velocity driving straight so that the transient responses are minimal at the start. Additionally it is desired that the vehicle be in zero-yaw attitude, meaning that the vehicle is driving straight. The driver is to maneuver the vehicle through the rst gate and then steer the vehicle into the targeted lane and through the second gate within the determined distance. This maneuver had to be attempted in a portion of road that is as straight and at as possible as to mitigate any non-accounted for e ects. The driver?s ability to negotiate the maneuver consistently will help insure that su cient data will be collected during the test series. The layout of the Single Lane Change is illustrated below in Figure 3.2. Figure 3.2: Single Lane Change Maneuver Diagram 48 3.3.3 Double Lane Change Object of Maneuver The double lane change is designed to capture the response of the vehicle in an obstacle avoidance maneuver, much like the single lane change. The main di erence in the two is that when executing the double lane change, the maneuver is not complete until the vehicle has returned to the lane that it was originally traveling in. This test is the most dangerous of the three as it incites the most lateral acceleration of them all and therefor leads to the most dangerous conditions for large, high center-of-gravity vehicles. The same measurements are of interest in the double lane change that were critical in the single lane change (path deviation, lateral acceleration, and settling distances) for each of the units. This maneuver stands to give the most data towards characterizing the rearward ampli cation as it is not just a single step change response nor a response in steady state, but instead a response to consecutive lane changes. Maneuver Path The path of the vehicle is approximately two consecutive single lane changes maneuvers. The vehicle is to start out in the outer lane and once it passes through the rst gate location. Change course and move to the inner lane before the next gate location. Once through the second gate location, there is a brief stretch of distance to be traveled in the inner lane, the end of this segment is marked by the third gate location. Once through the third gate location, the vehicle is to maneuver back into the outer lane once again before the fourth and nal gate location. Once the entire vehicle has passed through the fourth gate and has returned to a steady state operation the maneuver is complete. An illustration of the maneuver can be seen in Figure 3.3. 49 Figure 3.3: Double Lane Change Maneuver Diagram 3.4 Data Processing Once all data was collected, the raw data les were downloaded from the DAQ system and needed to be cleaned up and prepared for analysis. This process was broken down into a three step process: data handling (3.4.1), data formatting (3.4.2)), and real-time kinematic (RTK) corrections (3.4.3)). 3.4.1 Data Handling The computer that was mounted on the test vehicle recorded every run and stored the data internally. Following testing, all of the data sets were downloaded from the vehicle computer to a more powerful computer for the post-process work. A backup copy was left on the vehicle computer and another raw backup was stored in another location for redundancy. 3.4.2 Data Formatting Data recording was more e cient when the recorder ran continuously as the vehicle went around the track, with maneuvers on both the north and south straight segments. Data for each maneuver on the straight or curve were extracted from the larger le following the tests. This was accomplished by de ning zones around the track using known GPS coordinates for the track, as shown in Figure 3.4. 50 Figure 3.4: Zones for separating data recorded on the curves from data recorded on the straights. A MATLAB program was written to step through the large les line-by-line to de- termine the zone, and then to write all of the corresponding data for each zone to a new le. This program was also run against the individual recordings to ensure that all of the les representing a certain maneuver started and stopped at the same geographic point. The speed and maneuver were coded in the individual le names. An example of this is: ?NTRCI 35mph Dbl Ln Chng R2? which corresponds to the second running of a 56 km/h (35 mph) double lane change. The next step was coordinate transformations and unit conversions. The analysts chose to use the ISO vehicle coordinate system with positive z direction up, as indicated in Figure 3.5. Another MATLAB program was written to bring in the data from speci c runs and create an overlay plot so the accelerations of the tractor and three trailers could be compared. Figure 3.6 is the speci c case of angular rate about the z-axis in a double lane change. The four IMUs are listed in their order from front to rear in the combination vehicle. The upper graph is the raw data prior to the coordinate transformation, and the lower is after the transformation. The lower graph shows the four units of the vehicle yawing in their proper sequence . The data shown is after being put through a 10th order Butter-worth lter with a cut-o frequency of 3 Hz. 51 Figure 3.5: ISO standard frame-of-reference with Y to the left and Z up. Figure 3.6: Raw vs. Post Processed Angular Rates. After the necessary transforms were con rmed, a program was written in Python to it- erate through each le automatically and perform the rotations on the desired channels. The 52 program additionally converted the latitude and longitude coordinates to Universal Trans- verse Mercator (UTM) coordinates. This gave a set of rectilinear positioning coordinates of the tractor and trailers relative to an origin point close to the track. This simpli ed visualization of the data. 3.4.3 RTK Corrections In order to obtain global positioning coordinates, two GPS receivers were placed on each of the units in the combination. Novatel GPS receivers were on each of the units, while the lower cost alternatives, the u-blox GPS receivers, were on the tractor and rst trailer. The second and third trailer both carried an Oxford RT unit, but they did not function as discussed in Section 3.6. The accuracy (1 ) of the standard GPS positions that the receivers report are usually on the order of 1 or 2 meters (3 to 6 ft). This error comes from several factors, including atmospheric conditions, which is the largest contributor, as well as satellite and receiver clock biases, and noise in the measurements inside the GPS receiver. As the satellite?s signal propagates through the ionosphere and troposphere that surround the Earth, the electron dispersion and humidity can a ect the GPS signal. When several GPS receivers are operating within close proximity (several kilometers), these signal errors become highly correlated. Di erential GPS (DGPS) techniques take advantage of this fact and compensate for the correlated errors. DGPS methods can use the pseudorange, the carrier noise measurement, or a combination of both. Real-Time Kinematic (RTK) systems can take a precise carrier phase measurement and calculate global positions that are equally precise. The RTK system at NCAT uses a static base station with known coordinates near the track and compares its GPS measurements to those of a roving GPS receiver in close proximity. 53 3.5 Analysis of Experimental Data The experimental data was analyzed in order to look at the understeer characteristics of the LCV along with the body roll of each unit and axle steer. As described below in Section 3.6, there were some data collection issues that prohibited the accurate collection of acceleration data for the second and third trailers. However the second and third trailers there was adequate data collected on position, speed, and suspension de ections. This data was then compared to the results of the simulation which is described in Chapter 4. The vehicle characterization parameters mentioned above were obtained in conjunction with Western Michigan University as part of the collaborative e ort for the NTRCI project for LCV vehicle dynamics. That report can be found in the appendices of [6]. 3.5.1 Understeer Characteristics As described in Section 1.2.2 the understeer gradient e ects the vehicle?s ability to follow the path de ned by the steer angle of the steer axle tires or road wheel angle. The portion of the data that will contribute to the analysis of this characteristic is the data obtained during the constant radius maneuvers in the banked corners of the track. An overlay of two separate runs at separate speeds in shown below in Figure 3.7, this is to illustrate that the data at multiple speeds have the same nominal radii. Using the GPS data taken during the constant radius turns it was determined that the curve had a nominal radius of 476 ft. Steering Characteristics of the Tractor The understeer gradient was calculated using Equation (1.2) at various speeds using the experimental data. The measurement of the road wheel was taken from the string pot data and the lateral acceleration was both calculated analytically as well as recorded via IMU measurements. The equation for the analytical calculation of the lateral acceleration is seen in Equation (3.1). 54 Figure 3.7: Constant Radius Maneuver Path Overlay. ay = V 2 Rg (3.1) where ay = Lateral Acceleration V = Vehicle Velocity R = Turn Radius g = Gravitational Constant This equation is for vehicles that undergo constant radius turns on at surfaces which is not the case in this application. In order to correct for this, the gravitational contribution to the lateral acceleration is subtracted from the results from Equation (3.1). As described in Section 3.1, the corners in which the maneuvers took place had an 8a176 banking. So the resulting gravitational component of the lateral acceleration is g sin(8) = 4:478fts2 . In 55 addition to the equation for lateral acceleration, Equation (1.2) is listed below again for reference. K = L R 57:3 ay (1.2) where Fzo = Outer Vertical Tire Force Fzi = Inner Vertical Tire Force T = Track Width M = Vehicle Mass Ay = Lateral Acceleration hcg = Vertical Height of Center of Gravity (CG) There other series of equations that is of use are the equations for calculating the steer angles. Equations (3.2) - (3.4) show the calculations for the steer angle of both the inner and out wheels as well as the average steer angle, which is de ned as the Ackerman steer angle[16]. o = L(R +t=2) (3.2) i = L(R t=2) (3.3) = LR (3.4) where = Road Wheel Angle (deg) L = Tractor Wheel Base Length (steer axle to drive axle center-line) R = Turn Radius 56 These equations are set for low speed cornering where the lateral acceleration is negli- gible. In order to proceed with the calculations a value for both the road wheel steer angles and the hand wheel angle needed to be calculated from the measured data. In order to do this the data was broken down into two separate sections, the rst being the period the vehicle was transitioning from the straight-a-ways into the constant radius. This portion of the data is represented at the beginning and end of the data set. An example of the two separate sections from data taken at 35 mph can be seen in Figure 3.8. The second portion is the portion of the turn data that was analyzed as the constant radius section. It was this data that was tted with a linear regression to get best t linear curve, the resulting equation for the line is shown in Equation (3.5), and as can be seen by the low slope value of the section of the data is representative of a constant steering angle during the constant radius maneuver. y = 0:005x 3:2160 (3.5a) R2 = 0:00248 (3.5b) The same plot was generated, shown in Figure 3.9, at lower speeds to illustrate that the road wheel angle is indeed comparable. The di erences between the two can be attributed to high speed corrections made by the driver in an attempt to maintain the constant radius. The results of the calculations can be found in Table 3.1. Of note is that the under- steer gradient was calculated using only six discrete data sets at the various speeds of the maneuvers. 57 Figure 3.8: Sections of Constant Radius Maneuver. Figure 3.9: Sections of Constant Radius Maneuver. 58 Table 3.1: Understeer Gradient Calculations for LCV Tractor Target Vehicle Speed (mph) 20.0 25.0 30.0 35.0 40.0 45.0 Target Vehicle Speed (m/s) 8.94 11.17 13.41 15.65 17.88 20.12 Calculated Avg. Speed (m/s) 9.1558 11.0530 13.3027 15.1395 17.2976 19.2648 Measured Road Wheel Angle (deg) 2.3960 2.5726 2.6767 2.5444 2.2074 2.1626 Ackerman Angle (deg) 1.9799 1.9799 1.9799 1.9799 1.9799 1.9799 Under/Over Steer (deg) 0.4161 0.5927 0.6968 0.5654 0.2275 0.1827 Calculated ay (g) 0.0589 0.0859 0.1244 0.1611 0.2103 0.2609 ay Corrected for Banking (g) -0.0802 -0.0533 -0.0148 0.0220 0.0712 0.1217 Measured ay (g) -0.0130 0.0024 0.0372 0.0734 0.1289 0.1951 Understeer Gradient (deg/g) 7.0605 6.9006 5.6010 3.5035 1.0815 0.7003 Steering Characteristics of Trailer 1 The same calculations were performed on the data for the rst trailer in the combination. Those results are shown in Table 3.2. In the case of the rst trailer, the steer angle was measured by subtracting the measured heading of the 1st Trailer from that of the Tractor. In doing the above subtraction of heading angles the assumption was made that during the maneuver there was no slip for either of the two units. In other words the assumption is that the tires are pointed in their direction of travel [16]. This assumption can be made do to the low road wheel steer angle, should the steer angle increase above say 5 degrees then the assumption would lead to greater error then acceptable. An example of the output is shown in Figure 3.10. As illustrated the resulting steering input looks very similar to that of the Tractor. Steering Characteristics of Trailer 2 As described in Section 3.6, there were several issues with some of the data for the second and third trailer. As a result the recorded lateral accelerations were not reliable. For those purposes there is no value recorded for the measured lateral acceleration for neither the 2nd nor 3rd trailer. The results from the second trailer can be found in Table 3.3. 59 Figure 3.10: Tractor 1 Steering Input Table 3.2: Understeer Gradient Calculations for LCV Trailer 1 Target Vehicle Speed (mph) 20.0 25.0 30.0 35.0 40.0 45.0 Target Vehicle Speed (m/s) 8.94 11.17 13.41 15.65 17.88 20.12 Calculated Avg. Speed (m/s) 9.1372 11.0246 13.3071 14.9739 17.0491 18.7962 Measured Road Wheel Angle (deg) 2.4080 2.5350 1.5999 2.8151 2.6588 2.4725 Ackerman Angle (deg) 1.8513 1.8513 1.8513 1.8513 1.8513 1.8513 Under/Over Steer (deg) 0.5566 0.6837 -0.2514 0.9638 0.8075 0.6212 Calculated ay (g) 0.0587 0.0854 0.1245 0.1576 0.2043 0.2484 ay Corrected for Banking (g) -0.0805 -0.0537 -0.0147 0.0185 0.0652 0.1092 Measured ay (g) -0.0885 -0.0883 -0.0347 0.0119 0.0799 0.1367 Understeer Gradient (deg/g) 9.4858 8.0015 6.9825 6.1144 3.9518 2.5013 Steering Characteristics of Trailer 3 Similarly to the 2nd trailer the 3rd trailer also experienced data issues, because of this the data from the 3rd trailer was processed in the same manner that the 2nd trailer was and the outcome of the calculations is detailed in Table 3.4. 60 Table 3.3: Understeer Gradient Calculations for LCV Trailer 2 Target Vehicle Speed (mph) 20.0 25.0 30.0 35.0 40.0 45.0 Target Vehicle Speed (m/s) 8.94 11.17 13.41 15.65 17.88 20.12 Calculated Avg. Speed (m/s) 9.1307 11.0151 13.3039 15.1185 17.3635 19.3044 Measured Road Wheel Angle (deg) 1.6651 1.6311 1.5666 1.8866 1.9075 2.5102 Ackerman Angle (deg) 2.0897 2.0897 2.0897 2.0897 2.0897 2.0897 Under/Over Steer (deg) -0.4246 -0.4586 -0.5230 -0.2031 -0.1822 0.4206 Calculated ay (g) 0.0586 0.0853 0.1244 0.1607 0.2120 0.2620 ay Corrected for Banking (g) -0.0806 -0.0539 -0.0147 0.0215 0.0728 0.1228 Measured ay (g) n/a n/a n/a n/a n/a n/a Understeer Gradient (deg/g) -7.2440 -5.3767 -4.2034 -1.2639 -0.8595 1.6053 Table 3.4: Understeer Gradient Calculations for LCV Trailer 3 Target Vehicle Speed (mph) 20.0 25.0 30.0 35.0 40.0 45.0 Target Vehicle Speed (m/s) 8.94 11.17 13.41 15.65 17.88 20.12 Calculated Avg. Speed (m/s) 9.1372 11.0246 13.3071 14.9739 17.0491 18.7962 Measured Road Wheel Angle (deg) 2.4080 2.5350 1.5999 2.8151 2.6588 2.4725 Ackerman Angle (deg) 1.8513 1.8513 1.8513 1.8513 1.8513 1.8513 Under/Over Steer (deg) 0.5566 0.6837 -0.2514 0.9638 0.8075 0.6212 Calculated ay (g) 0.0587 0.0854 0.1245 0.1576 0.2043 0.2484 ay Corrected for Banking (g) -0.0805 -0.0537 -0.0147 0.0185 0.0652 0.1092 Measured ay (g) n/a n/a n/a n/a n/a n/a Understeer Gradient (deg/g) 9.4858 8.0015 6.9825 6.1144 3.9518 2.5013 3.5.2 Rearward Ampli cation As detailed in Section 1.2.2, rearward ampli cation at it?s simplest form it the tendency of a unit to exaggerate the inputs. In the case of LCV dynamics, this is most clearly illustrated in path deviation, lateral acceleration, and body roll of the vehicle. The lead unit for the LCV is the tractor and the inputs came from the driver and as those inputs are passed through the ?Train? they can grow in amplitude. The magnitude of ampli cation can be in uenced by several di erent factors. The most prominent ones are trailer length, dolly con guration, speed, and period of the maneuver [21]. The formal calculation of rearward ampli cation is de ned in Equation (3.6). 61 Rearward Ampli cation = Amplitude of Measured UnitAmplitude of Lead Unit (3.6) Should this value be greater than one, then the unit under measurement experiences a greater response to the inputs of the lead vehicle.There is no set characteristic that must be used to calculate the rearward ampli cation. Typically the value of lateral acceleration or yaw rate is used to determine the magnitude of the ampli cation. However, as described in Section 3.6, there were some complications with obtaining both of those measurements accurately for the latter half of the LCV. Due to that, the measurement of body roll was used in it?s place. Body roll was accurately captured for each vehicle with the use of string potentiometers attached to each unit?s ride axle. In the case of the tractor the potentiometers were installed both on the steer axle and on the second drive axle. The roll of the rst drive axle was assumed to be the same as the second. The input to the system is the hand wheel input by the driver. It was determined that the steering linkage had a 1=20 steering ratio, a plot of the input for a double lane change at 20 mph is illustrated in Figure 3.11. In each plot of the steering input there are eight di erent points of interest during the maneuver; those points are detailed below. POI 1 Start of the left turn to enter left lane. POI 2 Maximum left steer angle for initial lane change into left lane. POI 3 Point where hand wheel crosses from positive to negative after initial lane change. POI 4 Maximum right steer input for rst lane change. This is the corrective steer input to maintain the vehicle in the right lane after the rst lane change. POI 5 Maximum right steer input to return the vehicle to the original (right lane) lane of travel. POI 6 Point where the hand wheel crosses from negative to positive during nal lane change. 62 Figure 3.11: Double Lane Change Steering Input (20 mph) POI 7 Maximum left steer input for the nal lane change. This is the same corrective input that is seen at POI 4 but this is to return the right lane as opposed to enter the left lane. POI 8 Steering input returns to zero and maneuver is ended. Of note is that at higher speeds or shorter gate spacings POIs 4&5 are the same point on the input curve. At lower speeds or longer gate spacing there is a period of time where the driver is attempting to straighten out and steady the vehicle in the left lane. This results in two distinct peaks negative in magnitude which correlate to a right steer input. When the speeds are higher or spacing shorter there is no time for this correction and the end of the initial lane change coincides with the start of the second lane change. A plot of such input 63 is shown in Figure 3.12, notice the di erence between Figures 3.11 & 3.12 where there are two distinct negative peaks versus one. Figure 3.12: Double Lane Change Steering Input (40 mph) 3.5.3 Roll Behavior of LCV Units During 40 mph Double Lane Change With the knowledge that the steering input given by the driver is the overall input to the system the rst step is to analyze how the closest link responds to the driver?s input. The rst axle that responds to the input is the steer axle on the tractor. As it was one of the axles out tted with string potentiometers the body roll angle relative to the axle was captured during testing. A plot of the resulting roll for a double lance change at 40 mph can be seen in Figure 3.13. As illustrated, the body roll at the steer axle responds almost instantaneously to the input from the driver. This response of body roll is now the measurement that will be used to calculate the rearward ampli cation of all the subsequent LCV units. Keeping the analysis 64 Figure 3.13: Steering Axle Body Roll in the same order as the axles, the rst axle to investigate is the drive axle of the tractor. Any delays here would constitute the frame of the tractor was twisting during testing and causing a delay and /or ampli cation. A plot of the Steer axle roll compared to the drive axle roll can be seen in Figure 3.14. The same plot was generated for each of the towed trailers. That plot can be seen in Figure 3.15. As somewhat evident in the collection on plots in Figure 3.15, a slight delay and am- pli cation can be seen in the responses of the subsequent trailers. All of the signals were combined on the same plot and the outcome of that e ort is illustrated in Figure 3.16. As with some of the previous data sets it was necessary to lter the data in order to better analyze it as well as visualize it in plots. The signal for the body roll of all the axes except the steer axis were subjected to a low pass 10th order Butterworth lter with a cut-o 65 Figure 3.14: Drive Axle Body Roll frequency of 5 Hz. The new ltered responses were plotted again and produced Figure 3.17. As clearly shown in Figure 3.17 there is a de nite delay as well as a de nite increase in magnitude as the input is propagated down the LCV. In order to calculate an instantaneous value for the rearward ampli cation one point on the curve during the dynamics had to be chosen. For the following discussions and table the point that was chosen was the roll response to the highest input steer that was in right direction, POI 4& 5 in Figure 3.12. Using that point and the de nition of Rearward Ampli cation Equation (3.6), Table 3.5 was populated with the results from the calculations as well as the values used to plug into the equation. 66 Figure 3.15: Trailer Body Roll Table 3.5: Rearward Ampli cation for LCV during 40 mph Double Lane Change POI 4 Delta Rearward t rt r Ampli cation HandWheel 16.12 -2.1080 n/a n/a n/a Steer Axle 16.35 -0.7204 0.00 0.00 1.00 Drive Axle 16.84 -0.9271 0.49 -0.21 1.29 Trailer 1 17.08 -1.0980 0.73 -0.38 1.52 Trailer 2 17.60 -1.1510 1.25 -0.43 1.60 Trailer 3 18.05 -1.2310 1.70 -0.51 1.71 3.6 Data Quality Issues The rst speci c issue to arise was the attachment point for the string pot used to measure the horizontal displacement of the steer axle on the right side of the tractor had broken, so the string pot measured zero for the majority of the test week. Additionally it 67 Figure 3.16: LCV Body Roll (Un-Filtered) was discovered was that the messages from the engine CAN-bus were not logged. This was a result of code changing during the rst few days of the test week and the log statements that tell the computer which messages to be logged where not in the correct place in the con guration les for the DAQ. Noise was an issue raised for the string pots but this was not a result of anything other than process noise. Since the string pots were used to measure linear displacement they were put in a position as to track as much linear movement as possible for the axles; this meant a string pot was placed directly above and directly horizontal to the axle in its static position, shown in Figure 3.18. Due to the nature of the testing, recording noise-free data is not practically feasible. The string pot data re ects this with all the noise induced by the environment embedded 68 Figure 3.17: LCV Body Roll (Filtered) Figure 3.18: Lever arm mounting for string pots in each measurement. The string pots were mounted on 80/20 frames to the undercarriage of the vehicle in various positions. These axles are the rotation axis for the tires, which themselves experience a great deal of noise from the surface of the road. Not only does the 69 uneven pavement introduce noise to the measurements, but also the vibration of the test vehicle as it is driven down the road can be enough to a ect the data. In addition to the string pot issues, the two Oxford measurement units were used to record the positional and orientation data needed to fully characterize the motion of the two rear trailers of the test vehicle. An Oxford RT2500 was positioned on Trailer 2 and a RT3100 unit was located on Trailer 3 of the LCV. The RT units use both GPS and an inertial measurement unit (IMU) to measure the motion of the vehicle, and Kalman ltering is automatically applied by the RT system to fuse the data from these two types of sensor for optimal accuracy. Unfortunately, the devices were not con gured/initialized properly for the measurements and the RT data recorded were discovered to be erroneous. This was not uncovered until the data was thoroughly analyzed, as described below the initial data looked to be noisy but correct. Initially, it was observed that more noise was present in the signals than what had been expected or previously experienced. Filtering of some of the signals while the testing was being conducted showed results that were consistent with the expected motions of the vehicle for the maneuvers performed. As a result of various delays prior to testing and the resulting pressure to complete all testing within the planned time-frame, the testing had to be continued without a complete review and validation of the data integrity. 70 Chapter 4 Vehicle Simulation Results TruckSima174 is a simulation software that is developed by Mechanical Simulation Inc. The software is highly con gurable to allow for customization of the simulation to best match that of the desired vehicle. TruckSima174 is geared towards the heavy truck industry. This includes but is not limited to commercial vehicles, military vehicles, and large buses. TruckSima174 allows the user to input mass properties as well dynamic characteristics of the vehicle under test. These characteristics are normally entered as linear coe cients or tables of data. Once the vehicle has been con gured the user can then run the vehicle through a in nite number of sceneries. The user can either choose from pre-de ned maneuvers or can create their own. For this research the maneuvers that were undertaken during the experimental phase were mirrored in the simulation as closely as possible. Of note is that as packaged TruckSima174 cannot simulate the NCAT LCV as it can only simulate up to a double trailer con guration out of the box. In order to resolve this issue, a custom math package was purchased from Mechanical Simulation Inc. to allow for the simulation of the NCAT LCV. Out of the box with the custom solvers installed, TruckSima174 contains generic models of several di erent types of trailers to be implemented on the LCV. The default model was used as the starting point for the simulations. The data that was taken during the LCV characterization phase in conjunction with Western Michigan University was used to con gure the simulation model in order to more accurately represent that of the test vehicle. A screen shot of the TruckSima174 environment with the representative model can be seen in Figure 4.1. The LCV characterization report that was used to set up the model can be found in the appendices of [6]. 71 Figure 4.1: Screen Shot of TruckSima174 Environment 4.1 Vehicle Con guration As mentioned, the vehicle was con gured using the parameters that were determined during the characterization e ort, when measurements were not available the standard TruckSima174 parameters were left alone. The paths for each of the maneuvers described in Section 3.3 were entered in as closed-loop driver input paths. An example of this for the maneuver of a Double Lane Change is shown in Figure 4.2. The goal of this is to match the results of the simulation to that of the experimental in order to con rm that that simulation data is accurate and can therefore be used to simulate the vehicle in conditions that were not able to be run experimentally due to safety concerns. Complete details of the set-up of TruckSima174 can be found in Appendix A. Since the vehicle is a combination of six independent components it is unreasonable to expect the LCV to behave as a single vehicle. This means that when traveling in a straight line one should not expect all of the subsequent components to have zero for steer angles even if the lead unit does. What this means is that the driver must compensate for the di erent 72 Figure 4.2: Screen Shot of TruckSima174 Steering Path Input for Double Lane Change steer angles when attempting to drive in a straight line or along a speci ed path. The driver must constantly correct with slight turns of the steering wheel in order to maintain a straight path. This is a result of many minor contributions of the LCV components such as steering misalignment, suspension wear & misalignment, tire variations, and many other e ects. In an attempt to illustrate this, a simulation was run with the NCAT vehicle with no steering input at all. What this is indicative of is the driver removing their hands from the steering wheel and letting the vehicle settle at some steady state. The steering input from the hand wheel is shown in Figure 4.3. As shown, the steering input does not stay at zero as the vehicle settles to steady state. Additionally this is illustrated in the yaw of each of the units, shown in Figure 4.4. While the magnitude of the di erence is small it still stands to illustrate that the vehicle does not track perfectly straight. In all of the subsequent plots regarding the TruckSima174 data the nomenclature for the plots are detailed in Table 4.1. The response to the zero steering input serves another purpose other than illustrating the steady state characteristics of the LCV. It also serves to prove that the model is indeed stable and suitable for continuing the simulations. One crucial result of the simulation is 73 Figure 4.3: Steering Wheel Response to No Steering Input Figure 4.4: Yaw Response to No Steering Input the observation that it takes several seconds for each unit to settle after the start of the maneuver. For that reason extra time was added to the beginning of all the simulations to ensure that the vehicle was in steady-state prior to performing the maneuver. 74 Axle 1 Tractor Steer Axle Unit 1 Freightliner Tractor Axle 3 Rear Drive Axle of Tractor Unit 2 Trailer 1 Axle 4 Trailer 1 Rear Axle Unit 3 Converter Dolly Connecting Trailers 1 & 2 Unit 4 Trailer 2 Axle 6 Trailer 2 Rear Axle Unit 5 Converter Dolly Connecting Trailers 2 & 3 Unit 6 Trailer 3 Axle 8 Trailer 3 Rear Axle Table 4.1: Simulation Plot Key 4.2 Experimental Maneuver Simulation In order to simulate the maneuvers that were undertaken on the test track the steering inputs had to be placed into TruckSim. This was done by lling in tables of data with desired positions of the LCV and TruckSim would then steer the vehicle so that it would follow that path as best as it could. An example of such input is shown above in Figure 4.2. Each of the maneuvers and subsequent speeds that were undertaken on the test track were duplicated in the TruckSim environment. However, for the sake of space only the results from two of the speeds will be presented for each maneuver in this thesis. 4.2.1 Constant Radius Turn The constant radius maneuver is exactly what it sounds like, the vehicle travels through a constant radius curve at a constant speed. A screen shot of the animation of the LCV traveling trough the curve is shown in Figure 4.5. The path the vehicle traveled is shown in Figure 4.6. This path is the same for all speeds. Figures 4.7 & 4.8 show the roll of the vehicle at 25 & 40 mph respectively. Of note for the roll gures is that the roll is less for the 40 mph than the 20, this is due to the vehicle traveling at higher speeds and therefor having a greater lateral acceleration that counteracts the gravitational forces that are leading the vehicle to roll in the negative direction. Figure 4.9 illustrates the o -tracking of the LCV 75 during the constant radius turn. You can easily see that none of the units are actually able to follow the designed path and that each unit tracks to the inside of the previous unit in the combination. Figure 4.5: Screen Shot of Animation for Constant Radius Turn Figure 4.6: Path of LCV During Constant Radius Turn 76 Figure 4.7: Roll of Each Unit for 25 mph Constant Radius Turn Figure 4.8: Roll of Each Unit for 40 mph Constant Radius Turn 4.2.2 Single Lane Change The single lane change is simply maneuvering the vehicle from one lane to the other within the speci ed gate spacing. For this testing the gate spacing was set at 200 ft. The maneuver isn?t initiated until the vehicle is in a steady-state. The path is illustrated in Figure 4.10. In the constant radius turn the o -tracking that was evident was each unit traveling to the inside of the path set by the preceding vehicle. In the case of lane change 77 Figure 4.9: Unit O -Tracking During 40 mph Constant Radius Turn maneuvers, the same o set direction is seen at lower speeds. Yet as the speed increases the o set decreases, eventually becoming such that the trailing vehicle tracks outside of the leading vehicle. This is illustrated in Figures 4.11 & 4.13 where the rst is a zoomed section of of the single lane change at 25 mph and the second is a zoomed section of the single lane change at 45 mph. To illustrate the o -tracking focus should be paid to the rear axles of each unit, those being axle three and greater. The full version of the 45 mph single lane change is shown in Figure 4.12. The fact that the o -tracking switched from inside to out is due to the speed increase between maneuvers. The speed at which the o -tracking transitions is at 40 mph; below that the trailing axles track fully inside the lead axle whereas above that speed the axles begin to track outside the path of the lead axle. The same response can be seen in the plots of lateral acceleration as well as yaw. In both cases the highest values are that of the tractor and then decreasing down the LCV. Whereas at high speeds that behavior is reversed and the further back in the LCV the point of interest is the greater the response. This is the notion of rearward ampli cation. At lower speeds the rearward ampli cation is less than one, but after the vehicle crosses the critical speed limit the magnitude of the ampli cation becomes greater than one. That is to simply say the trailing vehicle exhibits 78 Figure 4.10: Path of LCV During 25 mph Single Lane Change Figure 4.11: Low Speed O -Tracking During 25 mph Single Lane Change a greater response than that of the leading vehicle. This is shown in the following gures; Figure 4.14 (Lateral Acceleration 25 mph) & Figure 4.15 (Lateral Acceleration 45 mph) & Figure 4.17 (Unit Yaw 25 mph) & lastly Figure 4.17 (Unit Yaw 45 mph). 79 Figure 4.12: Path of LCV During 45 mph Single Lane Change Figure 4.13: High Speed O -Tracking During 45 mph Single Lane Change 4.2.3 Double Lane Change The double lane change (DLC) is an obstacle avoidance simulation maneuver. The aim is to view the response of the vehicle to two quick subsequent lane changes that would occur in the case of the lane being suddenly obstructed and the driver having to rapidly adjust. The maneuver begins in the right lane and once the vehicle is at steady state the 80 Figure 4.14: Lateral Acceleration During 25 mph Single Lane Change Figure 4.15: Lateral Acceleration During 45 mph Single Lane Change driver steers the vehicle over to the left lane within a pre-de ned gate spacing. Once in the left lane the driver then returns to the right lane within the same gate spacing. A more detailed description of the maneuver was provided previously in Section 3.3. Most of the same characteristics are looked at for the double lane change that were examined for the single lane change. The di erences is that the double tends to excite larger ampli cations as 81 Figure 4.16: Unit Yaw During 25 mph Single Lane Change Figure 4.17: Unit Yaw During 45 mph Single Lane Change it is the culmination of two rapid lane changes instead of only one. The same gate spacing of 200 ft was used for the double that was utilized in the single lane change. The path of the double lane change is shown in Figure 4.18 (25 mph), the plot for 45 mph DLC is shown in Figure 4.19. 82 Figure 4.18: Path of LCV During 25 mph Double Lane Change Figure 4.19: Path of LCV During 45 mph Double Lane Change One of the di erences between the single and double lane changes is that due to the short distance that is traveled in the left lane prior to performing the 2nd lane change there is little to no o -tracking shown in the rst lane change regardless of speed. This is illustrated in Figures 4.20 & 4.21 which are for 25 and 45 mph respectively. 83 Figure 4.20: Low Speed O -Tracking During 25 mph Double Lane Change (1st Lane Change) Figure 4.21: High Speed O -Tracking During 45 mph Double Lane Change (1st Lane Change) However when examining the 2nd lane change you can discern the same experience with the o -tacking as illustrated in the single lane change. This is explained by understanding that the initial lane change is never fully completed as the vehicle is not allowed to settle prior to commencing the return lane change. To examine the o -tracking for the return lane change see Figures 4.22 & 4.23 again for 25 and 45 mph respectively. 84 Figure 4.22: Low Speed O -Tracking During 25 mph Double Lane Change (2nd Lane Change) Figure 4.23: High Speed O -Tracking During 45 mph Double Lane Change (2nd Lane Change) One of the most informative plots to look at for the double lane change is the lateral acceleration of each unit in the LCV. From this plot, not only can rearward ampli cation be examined but also settling time and stability of the units can be assessed. Figures 4.24 & 85 4.25 show the lateral acceleration according to TruckSim for the 25 and 45 mph double lane changes. Figure 4.24: Lateral Acceleration During 25 mph Double Lane Change Figure 4.25: Lateral Acceleration During 45 mph Double Lane Change As shown in the single lane change, the rearward ampli cation is less than one at the lower speed of 25 mph and greater than one at the higher speed of 45 mph. The more interesting element is the settling time of the third trailer as well as the amplitude of the 86 oscillation in the acceleration. At the lower speed of 25 mph the third trailer follows the acceleration patterns of the rest of the LCV fairly indistinguishably. Yet with an increase of only 15 mph, the third trailer becomes rather unstable and very oscillatory in it?s response. This is speci cally evident in the return lane change. This gross increase in lateral accelera- tion led to the simulation at highway speeds of 60 mph for exploratory purposes. That plot is shown in Figure 4.26. Figure 4.26: Lateral Acceleration During 60 mph Double Lane Change While it may be hard to discern what occurred during the maneuver, it is easily noted that the behavior of the third trailer is nothing like that of any other unit. The reason for this is that at the speed of 60 mph the simulation of the double lane change resulted in the third trailer rolling over in the middle of the return lane change, which is easily illustrated in the plot of unit roll for the maneuver which is shown in Figure 4.27. The third trailer of the LCV actually experiences wheel lift-o in the middle of the initial lane change and then when the tractor tries to navigate back into the original lane the third trailer rolls over and takes the second converter dolly with it. Another property of the 60 mph double lane change is how the tractor seems to be uninterrupted by the instability and subsequent crashing of the third trailer during the maneuver. 87 Figure 4.27: Unit Roll During 60 mph Double Lane Change 4.3 Experimental vs. Simulation In an attempt to compare the results from the TruckSim simulations and that of the data taken experimentally, several overlay plots were generated. The following plots represent the data collected and simulated for a 30 mph double lane change. The rst plot to examine is the driver input which is the hand wheel angle. That comparison is illustrated in Figure 4.28. There are several key di erences between the two data sets. The most noticeable is the early steer into the left lane in the experimental data. This can be explained by under- standing that the driver in the experimental phase had a much longer preview time than the simulation did. That is to say that the driver knew what was coming and started to turn to the left sooner than he should have. Whereas in the simulation, the computer does exactly what is laid out in the simulation and does not begin to make the turn until the exact moment the LCV passes through gate 1. This results in a lower steer input for both the initial departure from the original lane as well as the return into the original lane. Additional e ects of this discrepancy will be described shortly. Another key point of interest is the articulation angle of the two converter dollies. The experimental data was passed through a low pass lter to remove some of the noise so that 88 Figure 4.28: Steering Input Comparison for Simulation & Experimental Data the two data sets could be plotted together. The calculated and measured articulation angles are presented in Figure 4.29 & 4.30 for the rst and second dolly respectively for the 30 mph double lane change. As shown, the two data sets match up rather well with minimal disagreements. The early steer input can be seen in both plots as well. This measurement was of some concern as it was the combination of two di erent string potentiometers placed adjacent to one another and measuring the linear displacement of the proceeding trailer. That measurement was then converted to an angle based on the radius of the trailer tongue. This shows that the methodology and implementation of the articulation measurement was su cient in capturing the angle. This articulation angle serves as the steer input into the proceeding trailer, hence why from a dynamic analysis standpoint it is a crucial measurement to get right. The last set of comparison plots to present are that of unit roll. The roll for the rst and last trailers is presented below in Figure 4.31 & 4.32 respectively. While the two do not agree as much as would be desired the di erences are believed to be because of the longer 89 Figure 4.29: Dolly 1 Articulation Angle Comparison for Simulation & Experimental Data Figure 4.30: Dolly 2 Articulation Angle Comparison for Simulation & Experimental Data preview time described above. Since there was a lower steer rate input for the experimental data than that of the simulation data set it would be expected that the input would not 90 incite as much roll. Of interest is that for the steer input back into the original lane of travel the rst trailer comparison appears to match rather well. This is due to the preview time of the drive being shortened during the middle of the maneuver. In essence by the time the driver has successfully navigated the vehicle into the left lane, he has to begin his transition back into the original lane of travel. this causes an increased steer rate than that of the original lane change and therefore incites a greater roll response from the subsequent units. Figure 4.31: Trailer 1 Roll Comparison for Simulation & Experimental Data 91 Figure 4.32: Trailer 3 Roll Comparison for Simulation & Experimental Data 4.4 Double vs. Triple LCV One of the best ways to assess the safety of a triple is to compare it to a standard that is prominent currently. To that end, several simulations were run using a standard double semi-trailer con guration. The settings for the vehicle were all left as packaged from TruckSim with the assumption that they were representative of the average parameters for the vehicle. Three separate simulations were chosen for comparison, two of which were double lane change maneuvers and the third a single lane change. The double lane changes were run at 45, 55, and 60 mph. The single lane changes were compared at 45, 55, and 65 mph. For all of the maneuvers the characteristics of interest where roll, yaw, and lateral acceleration; each of those values will be plotted for comparison. 92 4.4.1 Single Lane Change Comparison The single lane change is a good starting point for assessing the stability of any vehicle. It is a low risk maneuver as it is not intended to excite large dynamic responses, especially in the case of the SLC set up for this research with a gate spacing of 200 ft. Figures 4.33 - 4.35 show the varying yaw responses for 45, 55, and 65 mph respectively. Figure 4.33: 45 mph Single Lane Change Double vs. Triple Yaw Comparison As illustrated in the gures, at lower speeds the units tend to behave identical. This is especially the case with the trailers. As the speeds increase though the responses begin to vary, most notably with respect to the tractor. Of note is that at the 65 mph speed the triple LCV experienced rollover of the third trailer, that is why the triple data plots end abruptly. One interesting characteristic that is illustrated is the lower oscillation of the tractor, this is a result of the heavier load in the triple. At the lower speeds the heavier load does not require as much correctional steering as the double. That correctional steering to keep the vehicle in the lane is the cause of the oscillation in the yaw of the double combination. The next characteristic to observe is the roll behavior of each vehicle during the same double 93 Figure 4.34: 55 mph Single Lane Change Double vs. Triple Yaw Comparison Figure 4.35: 65 mph Single Lane Change Double vs. Triple Yaw Comparison lane change maneuvers. Figures 4.36 - 4.38 show the varying roll responses for 45, 55, and 65 mph respectively. 94 Figure 4.36: 45 mph Single Lane Change Double vs. Triple Roll Comparison Figure 4.37: 55 mph Single Lane Change Double vs. Triple Roll Comparison The same heavier load that helped the triple in the yaw comparison hurts it in the roll comparison. Because of the heavier load, the vehicle experiences much larger roll than that 95 Figure 4.38: 65 mph Single Lane Change Double vs. Triple Roll Comparison of the double. This also causes the triple to ?rock? after each maneuver and takes some time to settle out. The roll of the vehicle is the most visual evident characteristic for large trucks with their innate high center of gravity. Of note is that none of these speeds were even attempted during the experimental phase as they were all beyond the driver?s comfort level. The visual of the triple settling out after the lane change is not one that would be comforting to any motorist. The nal characteristic that will be explored is the lateral acceleration of the units in the combinations. Those responses for 45, 55, & 65 mph are plotted below in Figures 4.39 - 4.41 respectively. The lateral acceleration response is somewhat a combination of both the Yaw response and the Roll response. The trailers appear to response identically at the lower speeds, as the speeds increase the triple begins to experience higher amplitude along with a shorter period of oscillation for the response. It can be seen in Figure 4.40 that the third trailer on the triple takes a considerable amount of time to settle after the maneuver. This is an indication 96 Figure 4.39: 45 mph Single Lane Change Double vs. Triple Lateral Acceleration Comparison Figure 4.40: 55 mph Single Lane Change Double vs. Triple Lateral Acceleration Comparison that the trailer is approaching it?s stability moment and any increase in speed could lead to rollover, which is exactly what happens at 60 mph. 97 Figure 4.41: 65 mph Single Lane Change Double vs. Triple Lateral Acceleration Comparison 4.4.2 Double Lane Change Comparison The next maneuver that was simulated for comparison was the Double Lane Change (DLC). This maneuver is meant to excite large dynamic responses and that is very evident at the higher speeds. Figures 4.42 - 4.44 show the varying yaw responses for 45, 55, and 60 mph respectively. The yaw responses of the vehicles behaves similar to that of the single lane change. At lower speeds the units, especially the tractors appear almost identical. Then as the speeds increase the tractor experiences a decrease in amplitude while the trailers appear to be una ected. That is, until you reach 60 mph at which the third trailer in the triple combination experiences rollover. The next characteristic to observe is the roll behavior of each vehicle during the same double lane change maneuvers. Figures 4.45 - 4.47 show the varying roll responses for 45, 55, and 60 mph respectively. The roll response of the triple especially does not look ideal at any speed. The third trailer fails to settle fully in any of the simulations. The responses of the triple truck and 98 Figure 4.42: 45 mph Double Lane Change Double vs. Triple Yaw Comparison Figure 4.43: 55 mph Double Lane Change Double vs. Triple Yaw Comparison rst trailer appear to match that of the double truck and rst trailer more closely during the double lane change than the single lane change. This is believed to be because once in the 99 Figure 4.44: 60 mph Double Lane Change Double vs. Triple Yaw Comparison Figure 4.45: 45 mph Double Lane Change Double vs. Triple Roll Comparison left lane the driver must already begin the maneuver to go back to the right lane, therefor not allowing any of the units to begin to settle in the left lane. The nal characteristic that 100 Figure 4.46: 55 mph Double Lane Change Double vs. Triple Roll Comparison Figure 4.47: 60 mph Double Lane Change Double vs. Triple Roll Comparison will be explored is the lateral acceleration of the units in the combinations. Those responses for 45, 55, & 60 mph are plotted below in Figures 4.48 - 4.50 respectively. 101 Figure 4.48: 45 mph Double Lane Change Double vs. Triple Lateral Acceleration Compari- son Figure 4.49: 55 mph Double Lane Change Double vs. Triple Lateral Acceleration Compari- son 102 Figure 4.50: 60 mph Double Lane Change Double vs. Triple Lateral Acceleration Compari- son The lateral acceleration of the units for the double lane change is not a good vote for the legalization of triple trailers on highways across the nation. The lateral acceleration the triple undergoes in the 45 mph maneuver alone is enough cause for concern. The double lane change is a maneuver that happens often and any vehicle that is going to be traveling at highway speeds needs to have the ability to perform such a maneuver in the event of an emergency. The triple LCV does not show any indication that it would be able to perform the double lane change at any speed greater than that of residential with any form of safety. This was also experienced in the experimental phase when it was attempted to perform the double lane change at 40 mph. After performing the maneuver the driver decided that it was too fast and only one attempt was made at that speed. 4.4.3 Double vs. Triple Conclusions After running the simulations and comparing the outputs it was interesting to see that for the most part the vehicle performed identical at lower speeds. This shows that the trailing 103 vehicles have little e ect on the leading vehicles from a dynamics standpoint at the lower speeds. However, as the speeds increased it was easy to see that the triple is much less stable and has a higher propensity for rollover making it unsafe. The triple experiences rollover in the double lane change maneuver at approximately 58 mph and at approximately 62 mph in the single lane change, both speeds that are well within the expected operating speed of any highway vehicle. It would appear that the triple needs much more safety integration before being suitable for the highway. 4.5 Sources of Error There were several large sources of error that were encountered in this project. the mitigation of such sources is key to improving the research and increasing the delity of the simulation. The main source of error for this phase of the research was human error in calculation the LCV characteristics. All of the distance measurements were ascertained by hand which innately leads to some margin of error. Additionally, material property assumptions had to be made in order to estimate both mass as well as compliance for the suspension system. Another source of error is the lack of complete characteristic data for the LCV as found at NCAT. As described in detail in Appendix A, there were a multitude of instances that characteristics were left as packaged from TruckSim. These characteristics might be a representative value of the average tractor and/or trailer they were not speci c to the NCAT LCV. It is unknown if there were any assumptions made in the creation of the simulation package for TruckSim that may be violated by the NCAT vehicle, which could serve as another source of error. 4.6 Improvements for Future Simulations As with any simulation, there is always room to improve upon it. This simulation is no di erent as there were several di erent areas in which the model was lacking to some extent. The most critical area in which the delity can be improved upon is the tire data. The tires 104 for the simulation were left as packaged from TruckSim. If the model is to be improved, the rst step suggested would be to obtain more accurate tire data and insert that data into the simulation. In addition to the tire data, more knowledge about the suspension kinematics would help serve the simulation. The data was also limited to what could be obtained by hand without the use large test equipment speci cally engineered to characterize heavy trucks. Some examples of such equipment would be a tilt table to better locate the CG of the units and a Kinematics and Compliance (K&C) measurement machine to better characterize the suspension. 105 Chapter 5 Conclusions This project was but a step in the direction of the research that needs to be completed to fully understand and characterize the dynamics of Longer Combination Vehicles. In conjunc- ture with experimental testing, an analytical model was developed and simulated through the use of a computer simulation package. The models created in the simulation phase are available for improvement and future use in parallel research. The known characteristics of LCV dynamics were shown to be present in both the experimental and simulation results. A method for capturing dynamic data on a LCV was both designed as well as tested. If the e orts were to be summed into one transfer function that would be Equation (5.1). The characteristic of most concern is that of the roll behavior of the third trailer as it reacts to the driver input of steer angle. This relationship is the basis for stability and subsequent safety assessments talked about in this chapter. TF = ThirdTrailerRollResponseDriverSteerInput (5.1) 5.1 Experimental Conclusions The experimental data concluded that the measurement of roll angle of the separate trailers was a reasonable predictor of rearward ampli cation of roll angle between the tractor and subsequent trailers. Additionally, the tractor and trailers showed understeer gradients that were expected given the mechanical linkage between the units. The custom designed DAQ worked adequately for the project, yet had it?s shortcomings. As described lateral acceleration data was not obtained for half of the LCV and as a result the traditional method for determining rearward ampli cation wasn?t possible. In future e orts, closer 106 considerations and precautions should be taken to ensure the correct data is being recorded and that the data recorded is reasonable in magnitude and shape. 5.2 Simulation Conclusions The simulation environment enabled multiple di erent scenarios being explored and allowed for the exploration of the behavior at higher speed unattainable during the ex- perimental phase. Many di erent maneuvers were simulated at many di erent speeds, for example through the iterative process it was determined that the critical speed for stability of the LCV during the double lane change was 58 mph. At speeds less than that the LCV would recover from the extensive dynamic response that was incited via the double lane change. Whereas at speeds above that, the vehicle would not recover and the third trailer would roll over during the returning lane change. This threshold is illustrated in Figures 5.1 & 5.2 where the roll of each unit is plotted versus time for 55 mph and 60 mph respectively. Figure 5.1: Unit Roll During 55 mph Double Lane Change In comparing the Triple LCV to a standard Double Trailer con guration it was shown that the triple had a lower stability limit in that it had a higher propensity for rollover at speeds that are to be expected of a highway vehicle. In addition, given the unsettling feeling 107 Figure 5.2: Unit Roll During 60 mph Double Lane Change most motorists have towards the double trailers it is hard to envision a triple trailer being received well if at all. 5.3 Final Conclusions Several of the dynamic characteristics were con rmed when comparing the experimental data to the output of the simulation. Measurements of all articulation angles along with vi- sual observations of the LCV (speci cally the third trailer) were in agreement. The rearward ampli cation of the unit roll and articulation were successfully measured and compared for single and double lane change maneuvers. Understeer properties were successfully measured using a constant radius turn. The combination of the above was adequate for a partial veri cation of the simulation. Taking all of this into consideration, the initial conclusions at this stage in the research is that LCV triples need a large number of safety equipment that is not yet available before they should be permitted on all highways throughout the US. Some of those improvements include a robust Electronic Stability Control (ESC) that can implement di erential braking on all of the units, as well the possibility of inducing torque on the fth wheel of each trailer in order to dampen out the oscillatory response seen in the simulations at higher 108 speeds. In order to achieve this, the communications between the trailers will need to be greatly improved. That improvement can be accomplished via a CAN bus that is common to the entire LCV. The additions that would be required would be more complete sensor packages that include but not limited to IMU?s, GPS receivers, and linear displacement measurements. The combination of the aforementioned sensors could be implemented into a control algorithm to predict and monitor the safety threshold for the unit and control the ESC accordingly to prevent any unwanted behavior. 5.4 Future Research The simulation model can be used to predict the potential impact of any changes in the con guration of the LCV. This includes changes in payload, dimensions, mechanical prop- erties, and many others. The vehicle used in this research was chosen because of both it?s availability and the availability of the testing course. It is to some level indicative of what a commercial LCV would be but that relationship could be improved upon. Most notably by payload weight and con guration. In addition the the payloads, the suspension characteris- tics of the units were assumed to be the same, if this research is to become complete that assumption has to be thrown out and each individual unit needs to be accurately modeled and tested. Perhaps the most in uential modi cation lies underneath the LCV. The tires are how the LCV interacts with the road, all of the forces that act between the ground and the LCV go through the tires. Due to this relationship, the tire properties are a crucial improvement in order to increase the delity of the model. There are two ways of improving those properties, either extensive experimental e orts or some propriety agreement with a tire manufacture. The simulations showed that the vehicle was unstable at highway speeds, a potential area of interest would be to investigate how to best increase the stable speed of the LCV through either weight distribution or some other mechanical property adjustment. This could be achieved rst in simulation prior to being implemented on any vehicle fro experimental 109 testing. An additional area of interest would be to incorporate the ESC system described above, or some subset of it in order to asses it?s performance in increasing the stability and safety of the LCV for the purpose of highway travel. 110 Bibliography [1] USDOT. (2009) Figure 3-4. permitted longer combination vehicles on the national high- way system: 2009. US Dept. of Transportation. [Online]. Available: http://ops.fhwa. dot.gov/freight/freight analysis/nat freight stats/docs/10facts gures/ gure3 4.htm [2] FHWA. Truck-tractor semitrailer combinations. [3] R. Ervin, \An overview of the dynamic performance properties of long truck combina- tions," 1984. [4] FHWA. Western uniformity scenerio analysis. [Online]. Available: http://www.fhwa. dot.gov/policy/otps/truck/wusr/wusr.pdf [5] C. Mallikarjunarao, \Tank trailer stability analysis," 1979. [6] W. B. E. B. C. C. R. C. J. C. M. C. R. H. M. K. T. L. M. M. J. P. J. P. C. P. A. S. D. W. Doug Pape, Michael Arant, \U32: Vehicle stability and dynamics longer combination vehicles nal report," National Transportation Research Center Inc., Tech. Rep., 2011. [7] IIHS. (2011) Fatality facts large truck 2011. International Institute of Highway Safety. [Online]. Available: http://www.iihs.org/research/fatality.aspx?topicName= Large-trucks [8] P. S. Fancher, \A factbook of the mechanical properties of the components for single- unit and articulated heavy trucks. phase i. nal report," 1986. [9] P. Nilsson, L. Laine, and B. Jacobson, \Performance characteristics for automated driv- ing of long heavy vehicle combinations evaluated in motion simulator." [10] D. Pape, M. Arant, D. Hall, S. Nelson, J. Petrolino, O. Franzese, H. Knee, N. Wood, S. Yeakel, R. Hathaway et al., \U02: Heavy truck rollover characterization (phase-a) nal report," 2009. [11] S. N. U. A. R. H. M. K. K. R. O. F. H. K. T. L. Doug Pape, Michael Arant, \U19: Heavy truck rollover characterization (phase-b) nal report," National Transportation Research Center Inc., Tech. Rep., October 2009. [12] T. LaClair, H. Knee, O. Franzese, M. Arant, R. Hathaway, M. Keil, D. Pape, and D. Rhoda, \U24: Heavy truck rollover characterization (phase c)," Tech. Rep., 2010. 111 [13] ISO, \Road vehicles { heavy commercial vehicle combinations and articulated buses { lateral stability test methods," International Organization for Standardization, Tech. Rep. ISO 14791:2000, 2000. [14] N. N. H. T. S. Administration). Fatility analysis reporting system. [15] C. MacAdam, M. Hagan, P. Fancher, C. Winkler, R. Ervin, J. Zhou, and S. Bogard, \Rearward ampli cation suppression (rams)," Tech. Rep., 2000. [16] T. D. Gillespie, Fundamentals of Vehicle Dynamics. Society of Automotive Engineers, Inc, 1992. [17] B. CHRISTOPHER, \Simpli ed analysis of the steady-state turning of complex vehi- cles," Vehicle System Dynamics, vol. 29, no. 3, pp. 141{180, 1998. [18] H. Yu, L. G uvenc, and U. Ozg uner, \Heavy duty vehicle rollover detection and active roll control," Vehicle System Dynamics, vol. 46, no. 6, pp. 451{470, 2008. [19] S. Zhou, L. Guo, and S. Zhang, \Vehicle yaw stability control and its integration with roll stability control," in Proc. Chinese Control and Decision Conf. CCDC 2008, 2008, pp. 3624{3629. [20] MOOS. Moos-ivp home page. Mission Oriented Operating Suite. [Online]. Available: http://oceanai.mit.edu/moos-ivp/pmwiki/pmwiki.php?n=Main.HomePage [21] P. Fancher and C. Winkler, \Directional performance issues in evaluation and design of articulated heavy vehicles," Vehicle System Dynamics, vol. 45, no. 7-8, pp. 607{647, 2007. 112 Appendix A TruckSima174 Con guration A.1 TruckSima174 General Overview TruckSim is a software package o ered by Mechanical Simulation. It is speci c to simulating and analyzing the dynamic behavior of medium to heavy trucks. Mechanical Simulation also produces CarSim, which is geared towards the more typical vehicle dynamics of smaller vehicles. The package provides the ability to run modeled vehicles through user- de ned tests and output the results in forms varying from tabular data, synchronous log les, plots, animations, and more. Data characterizing the vehicle to be modeled is entered into TruckSim in the form of mechanical properties, linear coe cients, or tables. As packaged, TruckSim cannot simulate an LCV triple; the most trailers that can be attached on the standard version is two. For that reason, a custom solver was purchased from Mechanical Simulation to enable to simulation of the LCV triple. Since it is a custom solution the set-up di ers slightly from the generic. Once the custom solvers were installed, the task of entering in the vehicle speci c data that represented the NCAT LCV was undertaken. A screen shot of TruckSim animating a simulation of the NCAT Triple can be seen in Figure A.1. A.2 Model Formulation The vehicle was modeled according to the measurements taken in conjuncture with Western Michigan University as part of an LCV characterization e ort. Where measurements were not available, the standard parameters in TruckSim were used. The following sections will describe in detail the steps taken to enter in the kinematic and dynamic properties of the NCAT LCV. Once installed and launched the rst screen that the user encounters is the 113 Figure A.1: Screen Shot of Simulation Animation main screen. That screen is shown in Figure A.2. In this window the user can select which vehicle con guration to use, which procedure to run, and what output if any they desire for the results. A description of all the options and why they are set that way is found in Table A.1. A.2.1 Vehicle Con guration Under vehicle con guration the user selects a vehicle class which in the case of the NCAT triple is S SS + S + dS + S + dS + S, each S represents an axle and each dS represents a converter dolly. Upon entering the con guration window for the vehicle the user is presented with the vehicle selection window, which is shown in Figure A.3. It is here that the user will select which vehicle components that are to be con gured to make up the vehicle under test. A description of the settings is shown in Table A.2 As seen, there are numerous di erent properties to be set. Of note is that for the NCAT LCV the second trailer is actually two trailers and a dolly which is why the vehicle type for 114 Figure A.2: TruckSim Main Window Table A.1: TruckSim Main Set-Up Screen Options Test Speci cations Vehicle Con guration Set the the data set of S SS + S + dS + S + dS + S to represent the custom solver for triple trailers. The vehicle selected is the custom vehicle built to match that of the test vehicle at NCAT, that process is detailed in Section A.2. Procedure Select which procedure is to be run. The set up for the procedures is detailedin Section A.4. Run Control Run Math Model The START button, will run simulation based on the selections made onpage. Output Variables Can write selected variables to external les such as .csv les or MATLAB.mat les. Results Animate Will run the animation and create a video of the simulation. Will not runif output variables are written to le. Plot Will plot the de ned plots selected in the procedure set up (Section A.4),can additionally de ne extra plots of interest. More Options Screen Options Can overwrite certain parameters de ned in the procedure set-up. Overlay Animations Can plot other simulations simultaneously as well as animate multiple sim-ulations at once. 115 Figure A.3: TruckSim Vehicle Window Table A.2: TruckSim Vehicle Con guration Options Lead Unit Vehicle Type The Tractor class is S SS as it is a single steer axle with double drive axles. Vehicle De nition Set to the con guration that resembles the NCAT Tractor, the detailedcon guration for the Tractor is described in section A.4. Trailer Trailer Type The selection here is a single axle trailer, which is what the single S repre-sents. Trailer De nition The Con guration that best represents the rst trailer in the combinationis selected, that con guration is detailed in Section A.2.3. Loads in First Trailer The payload is selected that represents the rst trailer, this too is detailedin Section A.2.3. Second Trailer Trailer Type For the second trailer the type dS + S + dS + S is selected to represent the remaining two trailers and converter dollies. This is selected as such because of the custom solvers that had to be installed in order to simulate a triple trailer. Trailer De nition The con guration for the Second & Third Trailer is outlined in the followingparagraph and accompanying pictures and tables. 116 Figure A.4: TruckSim Custom Second Trailer Con guration the second trailer is dS + S + dS + S instead of just dS + S for a typical double trailer con guration. In order to complete the con guration of the vehicle the user most con gure the custom solver options that are found once entering the con guration for the custom second trailer. Upon entering said con guration the screen encountered is shown in Figure A.4 and the accompanying description of options is outlined in Table A.3. A.2.2 Tractor Con guration The rst unit to be con gured is the tractor, upon exiting the custom con guration and returning to the main vehicle con guration screen (A.3)) the user can enter the con guration for the tractor. Once in the con guration the user sees the vehicle con guration screen shown here in Figure A.5. It is here that the mechanical properties of the tractor will be assigned and other properties de ned. A description of the settings is detailed in Table A.4. Tractor Sprung Mass Characteristics For the NCAT LCV the 2A Day Can Sprung Mass was selected as the vehicle type as that most closely re ects the actual tractor. The aerodynamics, animator shape, tires, 117 Table A.3: TruckSim Custom Solver Options Dataset Code This is left as packaged from TruckSim Custom Installation, additionally all of the comments in the parameter set boxes are untouched. Link 1 Link Type The selection here is a single axle trailer with a converter dolly, which is rep-resented by dS + S. Link De nition The Con guration that best represents the second trailer in the combinationis selected, that con guration is detailed in Section A.2.4. Link 2 Link Type The second link for the custom solver is the payload options of the secondtrailer. Link De nition The payload for the Second Trailer is de ned in Section A.2.4. Link 3 Link Type The selection here is a single axle trailer with a converter dolly, which is rep-resented by dS + S. Link De nition The Con guration that best represents the third trailer in the combination isselected, that con guration is detailed in Section A.2.4. Link 4 Link Type The fourth and nal link for the custom solver is the payload options of thethird trailer. Link De nition The payload for the Third Trailer is de ned in Section A.2.4. Figure A.5: Vehicle Con guration Window 118 Table A.4: TruckSim Tractor Set-Up Screen Options Sprung Mass Class & De nition Rigid Spring Mass; 2A Day Cab Sprung Mass. These selections were left aspackaged. Aerodynamics De nition Conv. Cab w/o Fairings 4.3 m Ref. Again selection was left as packaged asno information about the aerodynamics was obtained. Animator Shape De nition 3A Day Cab, selection was made to aesthetically represent the sprung massselection. Con guration was left as packaged. Tires De nition 3000 kg Steer, 3000 kg Drive/Tandem. Selection left as packaged. Steering Wheel De nition 1/25 (Typical) - Left as packaged Powertrain Class 6 x 4, axles 2 & 3 De nition 330 kW, 18 spd MT, 4WD - selection was left as packaged. Axle 1 Class Solid Axle (full K & C) Susp Kin, Con guration made to match that of the kinematics of the steer axle. Comp Con guration changed to mirror that of the NCAT tractor steer axle. Brakes Left as packaged Steering Selection was left as packaged Axle 2 & 3 Class Solid Axle (full K & C) Susp Kin, Con guration made to match that of the kinematics of the drive axles. Comp Con guration changed to mirror that of the NCAT tractor drive axles. Brakes Left as packaged Steering No steering on the drive axles Hitch De nition 5th Wheel - left as packaged. Position Dist Back (4760.9 mm) Y (0 mm) Height (1081 mm) 119 Figure A.6: Steer Axle Kinematic Con guration Window and steering wheel torque were left as packaged. The power-train was set at 6x4, axles 2 & 3 with the 330 kW, 18 spd, MT, 4WD transmission selected. The distances placed in all of the yellow boxes came from the LCV Characterization report. In the lower half of the window, the user is to select and con gure the suspension properties of the tractor the rst axle being the steering axle was con gured rst. The rst aspect to con gure was the suspension kinematics, this window is shown in Figure A.6. Again, the measurements that are placed in the yellow boxes are either re ective of the LCV Characterization Report or left as packaged in TruckSim. Tractor Steer Axle Suspension Characteristics Once back to the Vehicle Con guration Window, Figure A.5 the next set of properties to con gure are the suspension characteristics for the steer axle. That window is illustrated in Figure A.7. The rst property to set is the spring characteristics, based o the LCV Characteriza- tion Report the leaf springs where given a sti ness of 263N/mm at 2000N. The only other characteristic that wasn?t left as packages is the roll moment for the steer axle. That was 120 Figure A.7: Steer Axle Suspension Con guration Window con gured using data collected during the LCV characterization e ort. That screen is shown in Figure A.8. Tractor Drive Axle Suspension Characteristics The last component of the tractor to be con gured is the drive axles kinematics and suspension properties. The rst is the kinematic data, that screen is shown in Figure A.9. For the drive axles, the data from the characterization report was entered where appli- cable. In addition to the static data, the roll steer for the drive axle was not a simple linear coe cient as it was for the steer axle. Upon entering the con guration for the roll steer the user should see Figure A.10. The data for the roll steer was taken from the characterization report. This is the completion of the tractor con guration. A.2.3 Trailer 1 Con guration Once the con guration of the tractor is complete the next unit to con gure is the rst trailer. That screen is represented in Figure A.11, with the description of the elds in Table A.5. 121 Figure A.8: Steer Axle Roll Sti ness Con guration Window Figure A.9: Drive Axle Kinematic Con guration Window 122 Figure A.10: Drive Axle Roll Steer Window Figure A.11: Trailer Main Con guration Window 123 Table A.5: Trailer Con guration Options Sprung Mass 1A Trailer - Since the trailers were permanently loaded, measuring the sprung mass was not possible. Therefor the mass properties were adjusted using payload modi cations. Aerodynamics No Aerodynamics Animator Shape NCAT Box - Some minor adjustments were made to the default settings forcosmetic reasons, they bared no e ect on the simulation Dolly There is no dolly on the rst trailer as it attached to the fth wheel on thetractor. Tires Left as packaged as tire data was unavailable. Axle 1 Kinematics Con gured to match NCAT Trailer. Compliance Entered characterization report values. Brakes Left as Packaged Hitch Position De nition Trailer Kinematic Characteristics The rst con guration to complete for the trailer suspension is the kinematic data for the trailer axle, that con guration can be seen in Figure A.12. As with the other con gurations, the data was either left as packaged or taken from the LCV characterization report. Trailer Compliance Characteristics Next, the leaf springs needed characterization, that screen is shown in Figure A.13. The additional con guration of the roll moment is illustrated in Figure A.14. Trailer Payload Con guration The last step in con guring the rst trailer is to adjust the payload to match that of the loaded NCAT Trailer. The screen in which this is done is accessed through the main vehicle con guration window, Figure A.2, and is captured in Figure A.15. The numbers that were entered for the payload were taken from the vehicle characterization report and 124 Figure A.12: Trailer 1 Axle Kinematics Window Figure A.13: Trailer 1 Axle Leaf Suspension Con guration Window 125 Figure A.14: Trailer 1 Roll Sti ness Con guration Window were generated during the weighing of the loaded LCV axle by axle. This completes the con guration of the rst trailer. A.2.4 Trailers 2 & 3 Con guration The suspension of each trailer was con gured the same as that of the rst trailer since all of the trailers had identical suspensions and axles. The main di erence in the con guration of the second and third trailer versus the rst is the additional step(s) in con guring the dolly that accompanies the trailer. That process is detailed below in Section A.2.5. Seeing that the second and third trailers were identical from a kinematics and compliance standpoint, they were con gured exactly the same minus the variations in payload weight. That variation, along with the rst trailer is detailed in Table A.6. A.2.5 Con guring the Converter Dollies In order to con gure the converter dollies the user rst needs to ensure that all of the trailers are con gured correctly with respect to dollies and pintle hitches. In order for a trailer to ride on a converter dolly it must be speci ed in the trailer con guration that is 126 Figure A.15: Trailer 1 Payload Con guration Window Table A.6: NCAT Vehicle Gross Weight Measured TruckSim Measured TruckSim Unit kg kg lb lb Tractor & Trailer 1 32,986 33,029 72,900 72,994 Trailer 2 19,186 13,013 42,400 42,019 Trailer 3 18,552 18,588 41,000 41,080 Whole Vehicle 70,724 70,630 156,300 156,093 127 Table A.7: Trailer Linkage Con guration Dolly Pintle Hitch Fifth Wheel Unit Y/N Y/N Dist. Back (mm) Y (mm) Height(mm) Y/N Dist. Back (mm) Y (mm) Height(mm) Tractor N/A Yes 4760.9 0 1081 Trailer 1 No Yes 6692.9 0 774.70 No N/A Dolly 1 N/A Yes 1829 0 1100 Trailer 2 No Yes 6692.9 0 774.70 No N/A Dolly 2 N/A Yes 1829 0 1100 Trailer 3 Yes No N/A No N/A to use that dolly. That is to say for the second trailer that rides on the converter dolly between the rst and second trailer the dolly must be con gured within the second trailer con guration. The rst trailer is di erent because the fth wheel joint it rides on is part of the tractor and is included in the tractor con guration. Similarly, for the trailers that are to attach to a converter dolly (1st & 2nd) the pintle hitch must be de ned for the converter dolly to attach to. Table A.7 details which settings need to be con gured on each trailer such that the linkage is correct. All of the distances and heights came from the LCV characterization report. Once all of the linkages has been correctly con gured the last step in the model for- mulation is the con guration of the converter dolly compliance and kinematic data. Upon entering the dolly con guration screen the user should see Figure A.16. The settings of the converter dolly are detailed in Table A.8. Dolly Sprung Mass Characteristics The sprung mass con guration screen is shown in Figure A.17, the information was taken from the characterization report. Dolly Kinematics Characteristics 128 Figure A.16: Converter Dolly Con guration Window Table A.8: Converter Dolly Con guration Options General Options Sprung Mass NCAT 1A Dolly Animator Shape Dolly w/1A Tires Left as packaged as tire data was unavailable. Axle 1 X Dist. Back 2100 mm Kinematics Con gured to match NCAT Dolly. Compliance Entered characterization report values. Brakes Left as Packaged Hitch Position Dist. (1829 mm) Y (0 mm) H (1100 mm) De nition 5th Wheel Typical - As Packaged 129 Figure A.17: Converter Dolly Sprung Mass Con guration Window The kinematics con guration screen is shown in Figure A.18, the information was taken from the characterization report. Dolly Compliance Characteristics The suspension compliance con guration screen is shown in Figure A.19. The con gu- ration screen for the roll sti ness of the converter dollies is shown in Figure A.20. In both cases the information was taken from the characterization report. A.3 Model Stability An LCV should not be expected to move as a single unit in a straight line with zero steer angles in any of the constituent units. The driver must guide the vehicle straight with alternating right and left motions of the hand wheel to compensate for disturbance from wind and road. Tolerances and wear of the vehicle can cause misalignment to occur, requiring units of the vehicle to maintain a steer angle to keep an apparently steady state straight motion. This can be due to any minor di erences between the units such as steering misalignment, steering gear wear, suspension angle misalignment, frame damage, suspension 130 Figure A.18: Converter Dolly Kinematics Con guration Window Figure A.19: Converter Dolly Kinematics Con guration Window 131 Figure A.20: Converter Dolly Roll Sti ness Con guration Window wear, tire in ation variation, road crown, or even bearing drag. Measurement o sets can also be present in the test data as a result of imprecise alignment of sensors, imprecise calibration, or other systematic errors associated with the instrumentation setup. While it is impossible to pinpoint a cause for this behavior in the NCAT test vehicle, it was apparent in the test data. Of concern during analysis, TruckSim demonstrates a starting transient behavior that can be mitigated but not completely eliminated. This starting instability causes the simulation to start in a dynamic out-of-line orientation, which settles into a constant angle of articulation between the units of the LCV. This attitude continues for the duration of the simulation.FigureA.21 shows the yaw angles of each unit in the LCV during an un-steered simulation at 72 km/h (45 mph). This was simulated by setting the model in motion at a constant 72 km/h (45 mph) with an open loop steer controller having no steer angle de ned, which is as if the driver takes his hands o the steering wheel while maintaining the constant speed. This simulation was performed in order to demonstrate the stability of the model, but the results also show some interesting characteristics that are relevant to the transient maneuvers.. The results show transient dynamics during the rst few seconds. Afterwards, the vehicle stabilizes until a 132 Figure A.21: Unit Yaw During Open-Loop Maneuver dynamically neutral attitude is achieved, where the yaw angles of each unit maintain a near straight trajectory. This initial oscillation is typical of simulations as the masses \settle" from their initial conditions into equilibrium. Each maneuver begins with a short straight section during which the vehicle simulation is allowed to stabilize before the primary steering inputs begin. The steering wheel angle during this open-loop maneuver is plotted below in Figure A.22. A.4 Maneuver Paths & Con guration TruckSim came packaged with a multitude of default maneuvers as well as other inputs such as speed control pro les. For the sake of this research the speed controller was always set to constant as that was the goal of the maneuvers during the experimental phase. Each maneuver was set up as described below, when the user enters into the maneuver pro le screen they are presented with the screen shot shown in Figure A.23. The description of each setting and the subsequent selection is described in Table 133 Figure A.22: Steering Wheel Angle During Open-Loop Maneuver Figure A.23: Screen Shot of Maneuver Setup Screen 134 Table A.9: TruckSim Maneuver Set-Up Screen Options Driver Controls Target Speed Set to constant speed based on which maneuver it wasto represent Braking No braking needs as speed was to be constant Shifting Set to allow TruckSim to determine shift points usingan 18 speed transmission Steering Selected the driver input path needed for the desiredmaneuver Additional Data 3D Road Selected the 3D road con guration that was detailed fordesired maneuver Start and Stop Conditions Start Start at 0,0 for time and position on path Stop Stop at either 120 seconds or 350 m, whichever comes rst Plot De nitions De ne which plots that are desired when the data is plotted inside of TruckSim A.4.1 The Constant Radius Maneuver The constant radius maneuver was simulated by creating a circular path with two tan- gent roads that would slowly twist until the correct banking was achieved. The curve of the constant radius was created using the equation for a circle with a radius of 479 ft. The X-Y coordinates were calculated for the curve at every degree for the 180a176curve. Once this was created the points were shifted in order to become tangent with the two tangent road pro les that were created. A screen shot of the constant radius path is shown below in Figure A.24. As shown the path is two straight road segments that are tangent to the constant radius curve, this method was used over a complete circular path due to the orientation in which TruckSim would begin each maneuver. Many iterations and di ering attempts were used to get the LCV to orient on the 8a176banking and on the correct heading but none proved correct. Therefor the solution was reached that the vehicle was to start out on a at straight road that would twist from 0a176banking to 8a176banking over a span of 100 m. a graphical representation 135 Figure A.24: Constant Radius Path of the twist is shown in Figure A.25. The actual data points that were used for the twist are shown in Figure A.26. A.4.2 The Single Lane Change Maneuver TruckSim came pre-packaged with the single lane change maneuver, this was used as the base point to create the new maneuver. The points were generated to match the intended path outlined in Section 3.3. The resulting driver input path is shown in Figure A.27. The aim of the path was to travel down the centerline of the right lane and once through the rst gate steer into the left lane. In parallel with adjusting the default driver input path for the single lane change, the animation needed to be altered to match the desired path. This was done in a similar fashion that the driver path was determined. The di erence is that the points needed to represent the edges of the road instead of the centerline. Four sets of cones where placed to represent the maneuver, the middle two being the most crucial as the spacing had to match what was prescribed in Section 3.3. The resulting cone positions along with a graphical representation of the cones is shown in Figure A.28. 136 Figure A.25: 3D Representation of Gradual Lane Twist Figure A.26: Gradual Lane Twist 137 Figure A.27: Single Lane Change Driver Input Figure A.28: Single Lane Change Cone Positions 138 Figure A.29: Double Lane Change Driver Input A.4.3 The Double Lane Change Maneuver Similar to the single lane change, TruckSim already had a double lane change maneuver de ned so again it was modi ed to match the maneuver that was performed during the experimental phase. As detailed in Section 3.3, the driver was to steer back into the right lane after traveling through the second gate in the left lane. The same method that was used for de ning the single lane change path was used for the double lane change. The resulting driver input path is shown in Figure A.29. Again, the simulation cone positions had to be adjusted in order to match the newly de ned maneuver. The same cone positions where used from the single lane change however, there were two more sets of cones positioned in the right lane to represent the second lane change in the double lane change maneuver. In the case of the double lane change, the middle four sets of cones were critical as so to match the desired spacing of the gates outlined in Section 3.3. The resulting cone positions along with a graphical representation of the cones is shown in Figure A.30. 139 Figure A.30: Double Lane Change Cone Positions A.4.4 Future Potential Improvements for TruckSima174 Model As with any simulation, there is always room to improve upon it. This simulation is no di erent as there were several di erent areas in which the model was lacking to some extent. The most critical area in which the delity can be improved upon is the tire data. The tires for the simulation were left as packaged from TruckSim. If the model is to be improved, the rst step suggested would be to obtain more accurate tire data and insert that data into the simulation. In addition to the tire data, more knowledge about the suspension kinematics would help serve the simulation. The data was also limited to what could be obtained by hand without the use large test equipment speci cally engineered to characterize heavy trucks. Some examples of such equipment would be a tilt table to better locate the CG of the units and a Kinematics and Compliance (K&C) measurement machine to better characterize the suspension. 140