Adaption of Reachability Based Navigation Algorithms for off-road autonomous vehicles

Abstract

Autonomous vehicles offer a wide variety of benefits for the transportation sector, including increased efficiency, lower costs and wider access to transportation. Current research at the Automotive Research Center at the University of Michigan aims to extend autonomous capabilities to off-road vehicles. Military vehicles operating in a priori unknown environments stand to realis gains in combat effectiveness and personnel safety through enhanced autonomous capabilities. A number of approaches have been developed for path planning of autonomous vehicles, which select the vehicle's trajectory toward its objective. These algorithms must consider obstacle avoidance, path optimization under various metrics, and computational limits for embedded hardware. Reachability Based Trajectory design offers a particularly promising approach for mathematically provable safe navigation that can work with limited on-board real-time computational resources. However, existing implementations do not account for the unique dynamics of off-road vehicles, which are affected by variations of terrain, as well as tire-soil interactions. The focus of this work will be to adapt Reachability Based Trajectory Design algorithms to off-road applications. We will aim to extend the mathematically proved safety guarantees offered by Reachability Based Trajectory Design to be compatible with more complex dynamics models that capture the behavior of off-road vehicles in working in a priori unknown landscapes. We will work to factor in effects of terrain gradients, as well as tire-soil interactions to the algorithm, while ensuring feasible implementation with limited real-time processing capabilities.