Longitudinal research on autonomous vehicles incorporating BP neural networks and PID control models

Abstract

Addressing the limitations of existing dual PID longitudinal controllers for autonomous vehicles in terms of adaptability, this study innovatively proposes a new composite control model based on a BP neural network position PID and speed PID. First, a longitudinal control model for autonomous vehicles incorporating position PID and speed PID was constructed using MATLAB R2023a/Simulink simulation software, resulting in a dual PID control model. Subsequently, the position PID controller model and speed PID controller model integrated with the BP neural network were designed and implemented, forming a dual BP neural network PID control model. With the primary objective of tracking the desired path position, the BP neural network position PID controller and speed PID controller were integrated into the vehicle longitudinal dynamics model, constituting the new composite control model based on BP neural network position PID and speed PID. This was jointly implemented with the CarSim2019 software. From the simulation experiments, it can be seen that the new composite control model based on the BP neural network position PID and speed PID demonstrates excellent performance in tracking the target position. Specifically, compared with the traditional dual PID control model, the average error in position tracking accuracy of the composite control model is smaller, while significantly reducing the parameter tuning workload and further ensuring the vehicle’s position and speed tracking accuracy. Compared to the dual BP neural network PID control model, the average error in position accuracy is significantly reduced. Finally, target speed control experiments were conducted on an intelligent vehicle equipped with the Ubuntu open-source operating system. From the experimental results, it can be seen that compared to the traditional PID control model, the BP neural network PID control model achieves higher PID control accuracy while also improving the overall stability and accuracy of the system.