Dynamic Trajectory Planning for a Cross-Country Navigator

Abstract

Autonomous Cross-Country Navigation requires planning algorithms which supports rapid traversal of challenging terrain while maintaining vehicle safety. The central theme of the work is the implementation of a planning system which solves this problem. The planning system uses a recursive trajectory generation algorithm, which generates spatial trajectories and then heuristically modifies them to achieve safe paths around obstacles. Velocities along the spatial trajectory are then set to insure a dynamically stable traversal. Ongoing results are presented from the system as implemented on the NAVLAB II, an autonomous off-road vehicle at Carnegie Mellon University. The planning system was successful in achieving 5.1km autonomous test runs with obstacle avoidance on rugged natural terrain at speeds averaging 1.8 m/s. Runs of up to 0.3km at 4.5m/s were acheived when only checking for obstacles.