A Delaunay triangulation based node connection strategy for probabilistic roadmap planners

In this article, we revisit one of the key issues, i.e., node connection, in probabilistic roadmap (PRM) planners, which have been shown effective in high dimensional motion planning problems. We propose a new method of neighborhood selection strategy based on certain empirically observed properties of Delaunay triangulation of a random uniformly distributed point set. Our method allows a node in the network to have neighbors that are close to itself in the sense of Delaunay neighborhood. The algorithm introduced in this article is easy to implement and we show the boost in performance with the idea proposed in some preliminary experiments.