The Tangent Bundle RRT Algorithms for Constrained Motion Planning

摘要

A Rapidly-Exploring Random Tree (RRT) algorithm for path planning on constraint manifolds is presented. The basic idea behind the algorithm is to construct RRT on multiple tangent bundles which is approximations of the constraint manifold, and to perform lazy projections to the constraint manifold only for the nodes on the final path. We describe this Tangent Bundle RRT algorithm (TBRRT), and demonstrate the improved performance of our algorithm vis-a-vis existing methods through cases studies involving robot arms and mobile manipulators.