A Model-Predictive Urban Traffic Control Approach with a Modified Flow Model and Endpoint Penalties

摘要

Nowadays, congestion caused by traffic in urban areas is considered as a major problem. In order to make the best use of the existing road capacity, traffic-responsive control systems, including model-predictive controllers, are excellent choices. A model-predictive controller can minimize a cost function along a given time horizon. We propose a model-predictive control system that aims to reduce the congestion, and uses an internal flow model, which is our proposed modified version of the S-model. In the formulation of the objective function for the controller, we take into account the effect of those vehicles that remain in the network at the end of the prediction horizon until the network is completely evacuated. We formulate this effect as endpoint penalties for the MPC optimization problem. Finally, we will apply the designed controller to an urban traffic network and compare two scenarios, i.e., the fixed-time control case and the model-predictive control approach with the endpoint penalties proposed in this paper. The results prove the excellent performance of the model-predictive controller compared with the fixed-time controller.