ADAPTIVE CONTROL OF A LEGGED ROBOT USING AN ARTIFICAL NEURAL NETWORK by

Abstract

Results are presented on a neural network strategy for the control of a dynamic, locomotive system, in particular a one-legged hopping robot. The control task is to make corrections to the motion of the robot that serve to maintain a fixed level of energy (and minimize energy losses), which yields a stable periodic limit cycle in the system's state space. We effect the control of the robot by the use of an Artificial Neural Network (ANN) with a continuous learning memory. Previous applications of ANN in control systems have shown success for those systems that are to be controlled to a fixed point in the state space, however, they have not addressed the problem of control about more complicated motion in the systems state space, specifically about limit cycles. We investigate the design and simulation of an autonomous learning apparatus, aimed at devising a strategy for controlling a one-legged hopping robot. Through continuous reinforcement for past successes and failures, the control system develops a stable strategy for accomplishing the desired control objectives. Our results are presented in the form of computer simulations that demonstrate the ability of the ANN to devise proper control signals that will develop a stable hopping strategy, and hence a stable limit cycle in the robot's state space, using imprecise knowledge of both the current state and mathematical model of the robotic leg.