Constructing probabilistic roadmaps with powerful local planning and path optimization

This paper describes a new approach to probabilistic roadmap construction for path planning. The novel feature of the planner is that it uses a powerful local planner to produce highly connected roadmaps and path optimization to maintain the rapid query processing by a fast local operator. While most previous approaches obtain good roadmaps by advanced sampling methods, the presented approach concentrates on the method used to connect the samples. Empirical results show that the new approach outperforms the more traditional approach of using fast local planners in capability to produce roadmaps with only few connected components. Statistical analysis is used to identify features important for the efficiency of the local planners.