Why Probabilistic Roadmap Works?


Probabilistic Roadmap (PRM) is currently the most successful approach for path planning involving high DOFs robot. The main idea of PRM planning is to sample the robot's configuration space and construct a graph, called a roadmap, as a compact representation of the configuration space. The vertices of the roadmap are the sampled configurations, while an edge between two vertices represent a collision-free straight line path between the configurations that correspond to the vertices. Despite the success of PRM, little is known on the role of sampling distribution in PRM planning.

In this project, we study the role of sampling distribution in PRM planning. We articulate the distinction between sampling distribution and sampling source, which have been blurred in the literature. This study indicates that a suitable sampling distribution is critical for the success of probabilistic path planning because it represents the uncertainty that arises from our lack of information about the shape of the robot's configuration space.

Performance comparison of various PRM as the number of robot's DOFs goes up

Publications


  • . . . . . . Vol. . No. . pp. . ed. . ()