The adaptive control based on BP neural network identification for two-wheeled robot

For two-wheeled robot easily affected by uncertain factors, such as load change, the friction, road conditions and external affect, in the actual movement, this paper takes a model reference adaptive method based on BP neural network identification. Firstly, this paper designs a state observer for dynamic model by Kalman filter, which play a role of reference model. And then it separately uses two BP neural networks as identifier and controller, and online identifies nonlinear time-varying information of real system, to acquire the appropriate amount of each variable for motor control system, and to make the velocity of the actual system track the velocity of the reference model. This paper is through point stabilization control simulation experiments of wheeled robot, to test effectiveness of proposed method.