Study on fast model predictive controllers for large urban traffic networks

Traffic control is both an efficient and effective way to alleviate the traffic congestion in urban areas. Model predictive control (MPC) has advantages in controlling and coordinating urban traffic networks. But, the real-time computational complexity of MPC increases exponentially, when the network scale and the predictive time horizon grow. To improve the real-time feasibility of MPC, a simplified macroscopic urban traffic model is developed. Two MPC controllers are built based on the simplified model and a more detailed model. Simulation results of the two controllers show that the online optimization time is reduced dramatically by applying the simplified model, only losing a limited amount of control effectiveness. Additional techniques, like applying a control time horizon and an aggregation scheme, are implemented to reduce the computational complexity further. Simulation results show positive effects of these techniques.