Bipedal Locomotion Planning for a Humanoid Robot Supported by Arm Contacts Based on Geometrical Feasibility

Abstract

In this paper, we propose a multi-contact bipedal locomotion planning system for a humanoid robot, which can efficiently find feasible arm contacts and generate a quasistatic motion sequence supported by them. The key idea of our method is approximating the feasibility of an arm contact as its sustainability during a bipedal walking cycle, which we call “geometrical feasibility”. First, we plan a global path by RB-RRT with the reachability model and discretize it using footstep planning. In order to find sustainable contacts, we compute the intersections between the reachability volumes and environments along the planned footsteps, which are defined as contactable areas, and define contact candidates inside these intersections. Then, we prioritize possible contact sets based on their contactable areas and expected supportability. We pass these contact sets to whole-body motion planning process according to their priorities, and evaluate static equilibrium and kinematic constraints to generate quasi-static contact transitions. We apply the proposed system to four different simulation experiments, and conclude that it is a reasonable solution for the difficulty in predicting “feasible” arm contacts, which contributes to improving multi-contact capability of locomotion planning for a humanoid robot.