Randomized path planning for linkages with closed kinematic chains

Abstract

We extend randomized path planning algorithms to the case of articulated robots that have closed kinematic chains. This is an important class of problems, which includes applications such as manipulation planning using multiple open-chain manipulators that cooperatively grasp an object and planning for reconfigurable robots in which links might be arranged in a loop to ease manipulation or locomotion. Applications also exist in areas beyond robotics, including computer graphics, computational chemistry, and virtual prototyping. Such applications typically involve high degrees of freedom and a parameterization of the configurations that satisfy closure constraints is usually not available. We show how to implement key primitive operations of randomized path planners for general closed kinematics chains. These primitives include the generation of random free configurations and the generation of local paths. To demonstrate the feasibility of our primitives for general chains, we show their application to recently developed randomized planners and present computed results for high-dimensional problems.