Distributed Model Predictive Approach for Large-Scale Road Network Perimeter Control

In perimeter control, model predictive approaches have performed well in previous studies, but they were done in a centralized way. Considering real-time feasibility, extensibility, and functional stability, a distributed approach could be a promising solution particularly for cases with a large number of networks. In this paper, a new distributed model predictive approach is provided for large-scale road network perimeter control. In the model there are multiple local control agents which make their own decisions simultaneously at network boundaries after sharing necessary information. A control objective is provided upon the considerations of interaction between networks. Conditional weights are given for switching the control objective between greedy modes and cooperative modes. Case study results demonstrate that the proposed approach outperforms the fixed control and greedy control strategies.