|dc.description.abstract||A GPS based automated convoy strategy to duplicate the path of a lead vehicle is presented in this dissertation. Laser scanners and cameras are not used; all information available comes from GPS or inertial systems. An algorithm is detailed that uses GPS carrier phase measurements to determine relative position between two moving ground vehicles. Error analysis shows the accuracy is centimeter level. It is shown that the time to the first solution fix is dependent upon initial relative position accuracy, and that near instantaneous fixes can be realized if that accuracy is less than 20 centimeters. The relative positioning algorithm is then augmented with inertial measurement units to dead reckon through brief outages. Performance analysis of automotive and tactical grade units shows the twenty centimeter threshold can be maintained for only a few seconds with the automotive grade unit and for 14 seconds with the tactical unit.
Next, techniques to determine odometry information in vector form are discussed. Three methods are outlined: dead reckoning of inertial sensors, time differencing GPS carrier measurements to determine change in platform position, and aiding the time differenced carrier measurements with inertial measurements. Partial integration of a tactical grade inertial measurement unit provided the lowest error drift for the scenarios investigated, but the time differenced carrier phase approach provided the most cost feasible approach with similar accuracy. Finally, the relative position and odometry algorithms are used to generate a reference by which an automated following vehicle can replicate a lead vehicle's path of travel. The first method presented uses only the relative position information to determine a relative angle to the leader. Using the relative angle as a heading reference for a steering control causes the follower to drive at the lead vehicle, thereby creating a towing effect on the follower when both vehicles are in motion. Effective use of this method is limited to short following distances, or line of sight operation, similar to vision based following approaches. The following vehicle turns about a smaller radius than the lead vehicle, and this effect intensifies as following distance increases.
The second path duplication method allows for non line of sight operation by combining the vector odometry with the relative position to create a virtual leader to follow. The actual difference between the vehicles could be in excess of 100 meters, but the perceived distance is reduced to a predetermined value based on vehicle speed by re-generating the lead vehicle's position at a previous instance in time with the relative position and odometry information. Performance curves of path duplication accuracy versus following distance using different odometry techniques show that the partially integrated tactical unit provides the best performance, but the time differenced carrier approach offered very similar performance for a lower total system cost.
Both following methods were implemented on an unmanned ground vehicle. Tests showed following accuracy for the line of sight method was within 50 centimeters on straight sections, though the reference accuracy was centimeter level. The non line of sight method predicted the virtual leader position to within 5 centimeters for following distances ranging from 10 to 120 meters.||en