Research On Neural Network Control Method For Mobile Robot Trajectory Tracking Control

For the problem of mobile robot trajectory tracking control,this paper focuses on the designing of a method used as a trajectory tracking controller.In the method,it is combined with the logical reasoning ability of fuzzy control and the self-learning ability of neural network control.Under the control of the proposed controller,a mobile robot can independently and steadily track a given trajectory till to the given target point.The main work of this paper is as follows:First,we use intelligent control methods,so it is unnecessary to establish an accurate mathematical model for the controlled objects,which it is for the traditional control methods.By making full use of the flexible reasoning logic of fuzzy control system and the advantages of self-learning ability,nonlinearity and strong robustness of neural network control model,a mixed Pi-Sigma neural network(MPSNN)model is designed.It maps Takagi-Sugeno(T-S)fuzzy system to Pi-Sigma neural network structure to construct a MPSNN architecture.The reasoning process of T-S fuzzy system is used to explain the transformation process in the moving state of the mobile robot,and then gives a clear physical meaning to neural network.Further,the self-learning ability of Pi-Sigma neural network can modify T-S fuzzy rules and membership function online,avoiding the defect of the traditional T-S fuzzy system be limited to fixed fuzzy rules.Second,in order to speed up the optimization of the mixed Pi-Sigma neural network,a mixed Pi-Sigma neural network learning algorithm is designed based on the principle of error back propagation and gradient descent.Then,the MPSNN trajectory tracking controller is designed based on Pi-Sigma neural network.In the MPSNN,it is combined that the architecture and learning algorithm of the mixed Pi-Sigma neural network model.The proposed controller can track and control the angular velocity and linear speed of the mobile robot in real time during the movement,achieving accurate tracking of the robot's pose at any time.Finally,online simulation of the MPSNN trajectory tracking controller is implemented in MATLAB.Comparisons are made with the traditional Pi-Sigma neural network control method and the sliding mode control method.A series of experiments show that the proposed controller is able to track the given trajectory quickly and stably,reflecting its validity,superiority and robustness.