Controlling biped walking robots using genetic algorithms in mobile agent environment

Abstract

The design and implementation of a control program for biped walking robots using the genetic algorithms (GA) are presented. The most difficult problem with biped walking robots is that they have too many possible gaits. Generally it is impossible to find the optimal gait for a given route. In order to control biped walking robots, we have employed GA to determine the gaits of the robots. It is known that keeping the zero moment point (ZMP) in certain area is necessary for stable movement of a biped walking robot deriving ZMP is not theoretically difficult; it is just a matter of solving kinetic equations. The problem is that, for a certain series of ZMPs, the robot can have too many gaits and too hard to find the optimal one. We are using a control program using GA to produce approximately optimal gaits.