Simulation and control of biped locomotion-GA optimization

Abstract

The paper deals with the problem of biped locomotion control in the sagittal plane. A model of the biped with five linked rigid bodies is developed to simulate the system dynamics. This model has four joints of revolution, one at each link. A steady state genetic algorithm is used to find the necessary torques in each joint to obtain a desired trajectory for the biped's trunk center of mass. Each of the individuals in the GA consists of a set of command sequences given to the biped, the fitness being obtained through simulation of the resulting motion and its comparison to the desired trajectory.