The Applied Researsh Of Intelligent Predictive Control Algorithms For Inverted Pendulum

Inverted pendulum system is multivariable, nonlinear, strong-coupling and instability naturally,can effectively verify variety of control theory and methods, it is widely used in dustries such as military, aerospace, robotics and so on. Therefore, research of control algorithms based on the inverted pendulum has important theoretical and industrial applications.In this paper,we choose inverted pendulum system as object,combine intelligent control with predictive control to solve the existed problems of the control complexity and diversity of control objectives.Firstly,reviewed the research status of inverted pendulum home and abroad, the main control algorithm and general situation of intelligent predictive control development,do a simple description to the purpose and significance of the research,and points out the major work done.Secondly, working principle of the inverted pendulum system are introduced and the hardware circuit based on DSP are designed, the double linear inverted pendulum is studied, Selectivly analyze the method based on Lagrange equations mathematical model in introducing the current mathematical model.Through state feedback simulations of the pole-placement and LQR optimizing control methods,the disadvantages of these are pointed out. Thirdly,deeply study and research on the basis of BP neural network and generalized predictive control algorithm,the drawbacks of BP algorithm are point out, and improve it.Then a BP neural network predictive control algorithm is proposed, the model of a BP neural network prediction is established.Through simulation and identification achieve good results.Finally, the BP neural network predictive control algorithm is applied to double inverted pendulum system. The object model is built using a feedforward neural network while the control is realized through rolling optimization and feedback adjustment. The pendulum angle and displacement can be controlled simultaneously by the algorithm when it is applied to an inverted pendulum system. It is proved simulation result that the method can avoid the complicated mathematical analysis of the object and has the features such as fast conver-gence and strong robustness.