Manipulator motion planning with unconstrained end-effector under obstacle environment

Abstract

This paper proposes a novel way of carrying out manipulator motion planning with unconstrained end-effector (MMP-UE) under an obstacle environment, where an unconstrained end-effector implies that the end-effector can be displaced from the target object. We introduce two kinds of manipulator motion planning methods: the -based MMP-UE method and the -based MMP-UE method. The former finds a solution path for manipulator motion planning in the configuration space iteratively by extending the terminal trees in the constraint-satisfying subspace and a messenger tree in the configuration space and by connecting the terminal trees and the messenger tree to each other. The latter finds a solution path for manipulator motion planning in the task space by incrementally and randomly generating an extended point that can be added to the current master tree, finding the nearest point in the current master tree from the generated extended point, generating a set of reachable configurations that corresponds to the generated extended point, checking whether two consecutive sets of reachable configurations and are connectable, and finally extending the current slave and master tree if the consecutive sets of reachable configurations are connectable. We evaluated the performance of the two proposed methods in terms of the number of iterations, number of samples, number of generated nodes, and the total computation time through three different experiments: a toy train tunnel passing simulation, a wheel turning simulation, and a basket tunnel passing experiment using a real robot. The experimental results indicate that the two proposed methods succeed in finding the solution paths in all the experiments and that the -based MMP-UE method is faster and requires less storage than the -based MMP-UE method.