Abstract: For biped robot, to control the moving stability is the key technology for the design. The biped robot has a human figure and several pose when moving. It is difficult to control its dynamic stability using the traditional method. This article gives a dynamic energy pyramid method, derives the characteristic parameter ----overturning stability index that maintain stable, and gives the method to calculate the overturning stability index. According to the overturning stability index, using the fuzzy neural network adaptive control system can change the configuration and velocity of the robot real time to assure its overturning stability. It will be easier to control the stable movement of the robot using the parameter. The action generated by the system has good practicability and high reliability, also can reduce the danger in the movement of the robot.