Traffic signal control under stochastic traffic demand and vehicle turning via decentralized decomposition approaches

Abstract

Traffic congestion is a global pressing issue but can be mitigated via effective traffic signal control schemes. In this paper, based on a cell transmission model we coordinate the control of traffic signals at multiple intersections to maximize vehicle throughput on corridors or road networks, under stochastic traffic demand and vehicle turning. We formulate a two-stage stochastic mixed-integer linear program using finite samples of the uncertain parameter, and combine Benders decomposition with the alternating direction method of multipliers to develop spatially-temporally distributed algorithms for optimizing the problem. We test instances of traffic signal control on corridors and grid networks, generated based on synthetic and real-world traffic data. Our results show that (i) considering traffic uncertainty can significantly improve the signal control quality and (ii) decentralized decomposition approaches can quickly find high-quality signal plans for multiple intersections in complex road networks, and fully utilize the computation and communication technologies in smart-transportation infrastructures.