Hybrid probabilistic RoadMap - Monte Carlo motion planning for closed chain systems with spherical joints

We propose a hybrid probabilistic RoadMap - Monte Carlo (PRM-MC) motion planner developed under the general methodology of PRM. For a given robot, PRM planners generally need to sample and connect a large number of robot configurations in order to build a roadmap that reflects the properties (such as the connectivity or energy landscape) of the robot configuration space. The proposed PRM-MC planner uses Monte Carlo simulation to generate and connect neighboring robot configurations and uses PRM local planners to connect the connected components generated from MC simulation. This strategy follows the random sampling principle of PRM that leads to the probabilistic completeness of the PRM-type randomized planners, while exploring the continuity property of motion planning constraints to improve the computation efficiency and roadmap quality. We apply the PRM-MC approach to closed chain motion planning In this work. Our current planner uses rotation pivots as attempted Monte Carlo moves for 3D closed chains with spherical joints. Pivot motions are developed as an efficient way to deform closed chains without violating the closure constraints, which have proved problematical for randomized approaches. We discuss how to identify feasible rotation pivots of kinematic chains and utilize them in PRM-MC planning. Our simulation results show that the PRM-MC closed chain planner can build road maps with good connectivity and efficiently generate self-collision-free closure configurations for closed chain systems with many links and multiple loops.