The Research On Trajectory Following Control For Self-driving Car Based On Neural Network PID And Fuzzy Preview Theory

With the rapid economic development,the number of automobiles in cities has been increasing year by year.At the same time,due to differences in drivers’ driving skills,driving habits,control of road conditions,and physiological conditions,traffic accidents frequently occur.The advent of driving cars will solve this problem to a large extent.Trajectory tracking is a core problem of driverless cars.Its main content is to control the vehicle’s steering system(lateral path following control)and power drive system(longitudinal speed following control)so that the car follows the planned track.In the research process of the project,a Seven degree of freedom vehicle dynamics model and a nonlinear tire model were first established.Then,they were studied respectively in the longitudinal and lateral decoupling of vehicle dynamics.In the vehicle longitudinal control,a neural network PID controller with strong self-adaptive and self-learning capabilities was adopted.The vehicle lateral control was based on the optimal preview controller for feedback.At the same time of theoretical research of control methods,the establishment of vehicle and tire models were completed in the Matlab/Simulink platform,and the longitudinal and lateral controllers were simulated.The simulation results also verified that the controller can meet the requirements of unmanned vehicles.During the process of simulation verification of the lateral controller,when the vehicle speed was high and the curvature of the road changed greatly,the following effect of the lateral controller decreased significantly.In view of this,this paper further studied the Adjusted fuzzy system to dynamically update the preview time.Different preview time was selected under different vehicle speeds and road curvatures,which made the optimal predictive controller more adaptable,and the improved controller simulation was completed in the Matlab/Simulink platform.The comparison of simulation results shows that the improved controller has better path following ability than the optimal previewing controller with fixed preview time and the classical model predictive controller,and it can better reflect the driver’s driving process under various vehicle conditions and road conditions,which proved that the variable preview time controller has better vehicle control effect.