Speed control for a biped robot

Abstract

A biologically inspired model for a biped robot is developed. Each leg is modeled as a massless spring equipped with one radial and one angular actuator. The two legs in the model are attached to a point mass. The biped locomotion is modeled as a hybrid dynamic system that switches between four operation phases, with distinct dynamic behavior in each phase. A novel, simple yet robust control law that utilizes symmetry is developed to control the robot's speed, hopping height and balance during forward motion. The controller utilizes a feed forward artificial neural network, trained offline, to find approximate symmetric touchdown angles that are used by the control algorithm to determine actuator corrections. Simulation results show that the controller is able to maintain reference forward speed while maintaining balance during forward motion. Copyright © 2007 International Federation of Automatic Control All Rights Reserved.