Algorithm Research On Ball And Plate Control System Based On Vision

This paper selects ball and plate system GPB2001produced by Googol Technologycompany as the study object, to research control algorithms and theories. Ball and platesystem is a multi-variable and nonlinear system. This system’s control object is a plate drivenby two step motors. The plate’s movement can drive the ball’s movement, and then make it beable to balance at a specific location on the plate or to roll along a certain preset trajectory.First, this paper introduces the structure and working principle of the ball and platesystem, adopts Euler-Lagrange equation to derive and get the kinetic equation of the system.Because this system is nonlinear, sliding mode control (SMC) is adopted. Making full use ofSMC’s robustness and easy to implement and so on, this paper designs SMC controller basedon reaching law and equivalent control for ball and plate system, and simulates their effectthrough MATLAB. As SMC is easy to generate buffeting, hence radial basis function (RBF)neural network is used to weaken it. Combining with SMC and giving full play to thenonlinear mapping ability of the RBF neural network and its strong self-learning ability canachieve the adaptive sliding mode control of the ball and plate system. Finally basic particleswarm optimization (PSO) is introduced. As it is easy to trap into local minimum, this paperadds to a genetic factor to improve it. The improved PSO is used to optimize the center points,widths and weights of RBF neural network offline, to obtain better control effect. Equivalentsliding mode controller based on RBF optimized by PSO for ball and plate system ispresented. Simulation results through MATLAB before and after optimization are compared，and prove that the algorithm is feasible and work well.