CAD Model Based Autonomous Locomotion of Quadruped Robot by Using Correction of Trajectory Planning with RNN

Abstract

It is very difficult to make a mathematical model of the multi-leg locomotion robot and to design a motion control system for it, because of the robot's large-scale nonlinear property. In this paper we propose a new approach for dealing with such a large-scale nonlinear system and realizing the stable walking in various environments by using correction of trajectory planning with recurrent neural network (RNN) based on a CAD model. Firstly, a CAD model of a quadruped locomotion robot was developed using DADS/Plant, which is a software that treats multi-body dynamics. Then, the centralized hierarchical neural network (HNN) controller, which is used to make the robot realize a desired motion, was designed on the basis of this CAD model. Furthermore, on the basis of the same CAD model, the superior RNN controller, which is used to correct the trajectory planning of each joint of the robot, was utilized to solve such problems as the stability of the center of gravity of the robot body and the coordination of each robot foot during walking on an irregular terrain with obstacles and stairs. Consequently, the autonomous locomotion of the robot was realized. Simulations and experiments were performed to verify the effectiveness of the presented approach.