Design and implementation of balance control in a humanoid robot

Abstract

A proportional derivative control strategy was developed for the purpose of achieving balance in a humanoid robot. An artificial muscle model was adapted which modified physiological parameters for the purpose of controlling a lightweight robot skeleton. Gains were modified as a function of joint angles to permit low gain near the equilibrium point, and consequently to promote a human-like swaying behavior that is energy-efficient. The control strategy was testing by placing a non-zero initial condition on the ankle joint angle and observing the robot, both physically and in simulation, attempt to achieve a stable swaying pattern. This was achieved successfully in a simulation of the robot's mass and inertial parameters, but further efforts must be made to obtain the same behavior in the robot. The ability of a robot to successfully balance using a human-like sway pattern adds another successful biomimetic feature to humanoid robot control and in addition should improve the efficiency of such systems.