Dynamic Sampling RRT for Improved Performance in Large Environments

Basic RRT path planning algorithm uses a uniform random sampling over the entire free space of the environment in the process of producing new points for the RRT tree, which attain good results in cases of environments with simple and far apart obstacles. But it becomes much more challenging in large and complex environments with narrow passages between obstacles.This paper presents an improvement on the sampling method used in the rapidly-exploring random trees (RRT) search and path planning algorithm to improve its ability to explore in large, complex, and dense environments. The main idea is to make the sampling over a dynamic space that grows with the RRT tree, which increase the ability of exploring and finding narrow passages in large and complicated environments. A theoretical analysis of the fundamental principle is also presented with results from a simulated experiment to illustrate the benefits of the dynamic sampling.