Control of Traffic Signals in a Model Predictive Control Framework

This paper presents a signal control method of urban road traffic by using model predictive control (MPC), where online adjustments of all free parameters of traffic signals, i.e., cycle times, split times and offsets, are focused on simultaneously. A simple model of macroscopic traffic, whose dynamics are nonlinear involving binary representation corresponding to green and red signals, is formulated. The parameters of the model are calibrated by using data obtained from detailed microscopic simulation that yields realistic statistics. The simple model is transformed into a mixed logical dynamical system (MLDS), and a mixed integer optimization is applied in the MPC framework. The proposed control scheme makes the signals flexibly turn to red and green depending on traffic conditions in the road network. By adapting quickly to any changes in traffic conditions, the scheme generates appropriate traffic signals, and the traffic flow is kept optimum regardless of such changes. Results of optimization are also verified by microscopic traffic simulation using a detailed model. It is demonstrated that the proposed control scheme reduces the average number of vehicles and their average stay time in the network by outperforming other ones that control only the frequency and the phase of each traffic signal.