Development and Implementation of a Trajectory Prediction Methodology
Type of Degreethesis
MetadataShow full item record
Operation of unmanned aircraft in the United States' National Airspace System (NAS) is currently severely restricted, primarily due to the need to ensure adequate separation between manned and unmanned aircraft (UA). A particular problem in conflict avoidance algorithm is estimating where conflicting traffic is likely to be in the future. While most air traffic spends a large percentage of its time in straight and level flight, maneuvers are still quite common and must be considered. Incorporating uncertainty in tracking algorithms is well established, but current methods primarily only consider uncertainty related to sensor errors and modeling errors. They do not consider the uncertainty of pilot decisions regarding maneuvers. The objective of this research is to quantify the level of uncertainty in aircraft position due to pilot maneuvers and develop methods for incorporating that information into tracking and conflict avoidance algorithms. The uncertainty in position and velocity can arise from different sources such as sensor uncertainty, but a significant contributor is that the future behavior of non-cooperative aircraft is generally unknown. A pilot may maneuver for quite a number of different reasons. While aircraft on a cross-country trip will generally only make small course or altitude adjustments at various waypoints along their planned track, pilots that are just out "boring holes in the sky" or student pilots practicing various maneuvers may engage in fairly aggressive maneuvers unexpectedly. Thus it is helpful to quantify not only where a non-cooperative aircraft would be in the future given that it maintains its current velocity, but also where it could be if the pilot chooses to maneuver. In this study, time histories of aircraft tracks have been used to develop statistical models of aircraft maneuvers. Two sources of aircraft tracks have been used. Auburn University has a small fleet of flight training aircraft and GPS tracking devices were placed in these aircraft and their movements were tracked over approximately six weeks. Since these aircraft are used almost exclusively for flight training they represent aircraft that are most likely to maneuver. Each GPS unit was packaged with a high capacity NiMH battery pack to allow the unit to operate for up to a week without recharge. The second dataset was obtained from the FAA and includes tracks from aircraft operating over the contiguous United States. The Federal Aviation Administration (FAA) database include aircraft that are either operating under Instrument Flight Rules (IFR) or aircraft under Visual Flight Rules (VFR) but using radar flight following. IFR aircraft and VFR aircraft using flight following are typically traveling between two points so these aircraft would not be expected to execute many maneuvers enroute. These two dataset should provide a bound on the frequency of maneuvers. A statistical approach to analyzing the data was used to describe error in the projection due to maneuvers off the projected course. The aircraft tracking data was analyzed to determine how accurately the position of an aircraft could be projected forward in time assuming the aircraft travels at a constant velocity. At each point in time, the aircrafts' position and velocity were estimated using Kalman filter and other straight projection techniques. This position and velocity was projected forward over various time horizons and compared to the aircrafts actual position at that projected time. By accumulating the occurrence of error from the expected projection point, the confidence in projection both along- track and cross-track could be calculated for private, IFR and VFR aircraft. Frequency and extent of deviations for cooperative and non-cooperative air traffic can be used in testing conflict avoidance algorithms for unmanned aircraft. Confidence intervals were developed for compliant and non-compliant aircraft in the NAS at various flight levels in terminal and non- terminal environments.