A Gait Planning Method for Humanoid Robot to Step Over Discrete Terrain

Abstract

In order to ensure humanoid robot can cross obstacle autonomously in unstructured environment, a gait planning method is proposed for humanoid robot to step over 2D discrete terrain. According to the target step location and considering the geometric constraints, speed constraints, dynamic performance constraints and stability constraints of the robot in the real world, a differential evolution algorithm is used to search the feasible trajectories of the hip and ankle joints. In order to realize real time gait planning, robot joint trajectories for different target step locations is optimized offline and a gait library is established. By using Webots simulator in conjunction with MATLAB, the robot can send the data of obstacle detected by LiDAR in Webots to MATLAB to calculate corresponding step location and call joint trajectories in gait library automatically. The results show that the proposed method can realize stable walking of humanoid robot to step over stochastic discrete terrain.