Neural Network Control Method for Mobile Robot Trajectory Tracking
Abstract: In this paper, we study movement control problems of nonholonomic mobile robots trajectory tracking and propose an adaptive mixed Pi-Sigma neural network (MPSNN) control method combined effectively with logical reasoning ability of fuzzy control and self-learning ability of neural network control. This method maps Takagi-Sugeno (T-S) fuzzy system to Pi-Sigma neural network (PSNN) structure. It explains the motion state transition process for mobile robot with inference process of T-S fuzzy system and gives neural network certainly physical meaning. The backpropagation iterative algorithm of MPSNN is designed based on the principle of error back propagation and the gradient descent method. The self-learning ability of PSNN is used to adjust T-S fuzzy rules and membership functions on-line to make the trajectory tracking controller of the design have portability and adaptability. In addition, it also designed the quadratic interpolation method to dynamically adjust learning rates in the network and improve the error convergence efficiency. Finally, we design two MPSNN trajectory tracking controllers based on Pi-Sigma neural network and verify the validity and superiority of the proposed method and the designed controller by using MATLAB numerical simulation.
Loading