Freeway ramp PID controller regulated by BP neural network

A parameter adjustment method of PID controller with BP neural network is developed and applied to freeway on-ramp metering in this paper. Firstly, the objective of ramp metering is determined, and a traffic flow model to describe the freeway flow process is built. Then the learning algorithm of BP neural network for adjusting the proportional, integral and differential coefficients is formulated in detail. Based on the traffic flow model and in conjunction with nonlinear feedback theory, an on-ramp PID controller regulated by BP neural network is designed. According to real-time traffic status, BP neural network is used to adjust the PID parameters dynamically in order to minimize the performance index defined in terms of the density tracking errors. Finally, the controller is simulated in MATLAB software. The results show that the controller designed has good dynamic and steady-state performance. It can achieve a desired traffic density along the mainline of a freeway and thus avoid traffic congestion. This approach is quite effective to freeway on-ramp metering.