Dynamic Walking of Eight DOF Biped using Neural Net Controller

摘要

This paper deals with the design of a Neural Controller for dynamic walking of an eight degree of freedom biped robot. An eight DOF biped has very complex system dynamics. It has been shown that the neural controller gives much better performance than the conventional PD controller when the plant dynamic is complex and disturbances cannot be modeled. A two-layer neural net (NN) controller with standard back propagation weight tuning has been proposed in this paper for dynamic walking of the biped. Quantitative comparison shows the superiority of NN controller over PD controller.