Fast Multi-Contact Motion Planning Based on Best-Neighbor Search of Contact Sequences

Abstract

This paper presents a computationally efficient method for planning dynamic multi-contact motion. A low-dimensional dynamical model of a robot equipped with multiple contact points is developed, and a motion planning problem is formulated as a optimal control problem with continuous and discrete variables. An extension to the differential dynamic programming (DDP) framework enables efficient computation of optimal cost for a large number of locally modified contact sequences. Based on this finding, a novel algorithm for multi-contact motion planning is developed, and its performance is evaluated in simulations.