Anytime dynamic exploring rapid random tree approach in higher dimension search space for non-holonomic robotics

Abstract

With the breakthrough of exploring Rapid Random Tree and several other improvement efforts, the sampling-based motion planning method has been gaining ground in 2D planning and gradually being accepted by many systems as their global planning algorithm. There are many recent approaches on integrating the incremental nature and algorithmic simplicity of sampling-based approach with the agile replanning strategy studied on Incremental Heuristic based motion planning algorithms. Therefore, we develop an anytime dynamic RRT* for non-holonomic systems. We also implement Lyapunov function, which better represents the true-cost-to-go, in order to generate ideal and smoother trajectories. Our method proved to be more robust than some state-of-the-art planning algorithms, with lower cost and smoother path. We evaluate our algorithm with 3 major benchmarks in simulated as well as real-environment. We compared our algorithm with other major planning approaches and proved the cost yields between 8.5%~16.7% less cost and as may reach as low as 58.17% and 95% less than RRT and RRT* respectively.