Anytime Computation of Time-Optimal Off-Road Vehicle Maneuvers using the RRT*

Abstract

Incremental sampling-based motion planning algorithms such as the Rapidly-exploring Random Trees (RRTs) have been successful in efficiently solving computationally challenging motion planning problems involving complex dynamical systems. A recently proposed algorithm, called the RRT*, also provides asymptotic optimality guarantees, i.e., almost-sure convergence to optimal trajectories (which the RRT algorithm lacked) while maintaining the computational efficiency of the RRT algorithm. In this paper, time-optimal maneuvers for a high-speed off-road vehicle taking tight turns on a loose surface are studied using the RRT* algorithm. Our simulation results show that the aggressive skidding maneuver, usually called the trail-braking maneuver, naturally emerges from the RRT* algorithm as the minimum-time trajectory. Along the way, we extend the RRT* algorithm to handle complex dynamical systems, such as those that are described by nonlinear differential equations and involve high-dimensional state spaces, which may be of independent interest. We also exploit the RRT* as an anytime computation framework for nonlinear optimization problems.