Coupled Linear Programming Approach for Decentralized Control of Urban Traffic

Management of urban traffic systems is a challenging task, partly because queue spillbacks, which bring about loss of intersection capacities and lead to systemwide congestion, arise easily in moderately congested networks. One possible way of ensuring mobility is through proactive spatial distribution of queues to contain their impact locally. In this paper, a traffic signal control strategy based on this idea is proposed. The strategy is traffic adaptive and operates in a decentralized fashion. It is formulated as a coupled system of linear programs, each optimizing a local queuing pattern according to real-time queue information and history of boundary flows. The proposed control is constructed to be scalable and robust to system uncertainties. Its properties (e.g., avoidance of queue spillback and stabilization of traffic) are demonstrated through numerical experiments.