Development of path tracking software for an autonomous steered-wheeled robotic vehicle and its trailer.

摘要

Wright Laboratory, at Tyndall AFB, Florida, has contracted the University of Florida to develop autonomous navigation systems for a variety of robotic vehicles, capable of performing tasks associated with the location and removal of bombs and mines. One of the tasks involves surveying plots of land for buried munitions. The navigation subtasks include vehicle positioning, path planning, path following, and obstacle avoidance.; The author's master's research involved implementing, on a navigation test vehicle (NTV), a path planner that generates the shortest collision-free path between two vehicle configurations. Subsequent to this work, a second planner was implemented that generates an efficient route for surveying a polygon shaped field. These routes ensure 100% of the surface area is covered, provided that the path is accurately followed.; Accuracy in path following is critical to the task. There are hundreds of acres that currently require surveying. The sites are typically divided into regions, where each mission can take up to 4.5 hours. These sites are usually surveyed using parallel rows. By improving the accuracy of path following, the distance between the rows can be increased to nearly the detection width of the ground penetrating sensors.; There are two ways to improve the accuracy of path following: improve vehicle positioning and reduce control errors. The NTV is currently capable of calculating its position in real-time, accurate to within 6 cm on an average. The author's doctoral research involves developing a strategy that takes advantage of an accurate positioning system and ensures accurate path following for any steered-wheeled vehicle over a range of speeds (0-4.5 m/s).; This strategy involves a process for standardizing and smoothing a given path, calculating a target position in real-time, combining the desirable features of a high-level PID and adaptive pure pursuit steering controller, autonomously calibrating the steering feedback data, autonomously tuning control parameters, and scheduling the control gains as a function of the vehicle speed. Auto-tuning is accomplished with a relay tuner and a golden section search (GSS) optimization. This strategy was demonstrated in simulation and implemented on the NTV.