The 22nd COTA International Conference of Transportation Professionals (CICTP 2022)

This paper proposes a network traffic control method to enhance model-based traffic control with data-driven online adaptive optimization. A macroscopic traffic flow model is first developed for the model prediction. Then, the model is further enhanced in real-time based on the performance measurements using the adaptive optimization or learning algorithm. Integrating the data-drive optimization into the model-based predictive control, the proposed control method is able to identify the key model parameters and optimize the actual network performance. Experiments on a toy network are conducted to test the efficiency of the proposed control method. Simulation results show that the proposed method generates better control performance than the general model-based control method.

Traffic signal control, traffic flow model, model-based traffic control, adaptive optimization