IMPROVING RRT-CONNECT APPROACH FOR OPTIMAL PATH PLANNING BY UTILIZING PRIOR INFORMATION

Yuanshan Lin, Di Wu, Xin Wang, Xiukun Wang, and Shunde Gao

Keywords

Path planning, rapidly exploring random tree (RRT), optimal path, sampling-based algorithm, sampling strategy

Abstract

This paper presents a novel efficient path planning approach denoted as RRT-Connect++ for high dimension problems with differential constraints. This work focuses on obtaining sub-optimal path within short time, while most conventional approaches strive to quickly find a feasible path or improve the quality of a path at the cost of expensive planning time. The fundamental idea of this approach is to utilize prior information to guide the search. Three modifications on the original RRT-Connect algorithm are made: constructing sampling pools with those promising vertices of trees and picking random state from them; avoiding sampling from the explored regions; adding the middle vertices during the connection operation and testing regression of vertices to guarantee the quality of trees. The performance is compared with those of several other RRT-based algorithms with three experiments to demonstrate the quality of path returned by it and its planning time efficiency. Results from the three simulation experiments show that the RRT-Connect++ can quickly find higher quality path and its efficiency is higher as well.

Important Links:

Go Back