Foothold Selection Considering Constraint and Slippage Evaluation for Legged Robots

Abstract

Legged robots over a rugged terrain can be accomplished by taking the environment and control constraint models into account when planning foothold and robot movements. In this paper, the foot-terrain slippage evaluation model for 3-DOF leg based on dynamics model is proposed. The authors evaluate the geometrical characteristics of each cell on the local elevation map, checks slippage state constraints and kinematic constraints. Then, we employ reinforcement learning and imitation learning to develop the relationship between the local elevation map, robot state, and robot behavior. The method can help to select the foothold considered geometrical characteristics on the elevation map and the robot state can be determined. The experiments were carried out on legged robots walking over rough terrain in both simulation and real robotic platforms. And the experimental results have demonstrated the effectiveness of the proposed method.