Walking Simulation of Biped Robot on Inclined Plane Based on Gait Recognition

Abstract

Limited by the structure, the walking motion of the biped robot on the slope is prone to the problem of the centroid shifting back and forth, which causes the robot to fall. If the slope is known, we can achieve the purpose of centroid adjustment and stabilize the robot by compensating the robot's ankle joint based on the slope. But the robot's ability to perceive the environment is limited, it's hard for the robot to recognize the slop and then make effective compensation for the ankle joint of the robot. In this paper, we propose to use the idea of imitation and use people's ability to perceive the slope to complete the slope angle perception. Using Kinect sensors to obtain the walking information of the human body under different slopes, and then using LSTM neural network to jointly learn the degrees of freedom of multiple lower limb joints, complete the classification and recognition of different slopes, and then compensate the robot's ankle joints based on the slope inclination. On the WEBOTS simulation platform, the biped robot NAO is driven to reproduce the walking action in a slope environment according to the corresponding movement strategy and the label compensation method obtained according to the classification. The simulation results show that the NAO robot can complete the stable walking movement simulation under the corresponding gradient environment.