Control of Quadruped Locomotion Robot Using 2DOF Control System with Adaptive Learning Based on Feedback-Error and Learning.

Abstract

Generally speaking, it is clear that the servo system for motion control, such as in a locomotion robot, needs feedforward control in addition to feedback control to improve the control performance in the transient tracking responses. This is called a two-degree-of-freedom (2DOF) control system. However, this control strategy is based on the mathematical model of the controlled object. On the other hand, a quadruped locomotion robot is a typical MIMO system and a strongly nonlinear system. Thus, it is very difficult to create an exact mathematical model for control system design. We believe that most research work on quardruped locomotion robots uses local feedback loops with PD control. We propose a new control strategy for a quadruped locomotion robot using the 2DOF control system with adaptive learning based on feedback-error and learning. This system consists of a feedback control system based on eight independent local PD feedbacks as a decentralized controller and the feedforward control system of the neural network with feedback-error and learning as the centralized controller. We have succeeded in conducting the actual locomotion test with very smooths and fast locomotion using this control strategy.