Scalable real-time control of intersections in urban traffic networks thanks to a leader-follower coordination approach*

This paper proposes a distributed coordination mechanism between local feedback controllers that adapt the switching times of traffic lights at neighboring intersections. Some heavily loaded intersections are designated “leader intersections”. There, a local control agent sets switching times to minimize the expected queues, knowing vehicle passing times at local upstream sensors only, and sends its intended switching plan to neighboring “follower intersections”. Each follower agent uses this information to try to minimize the waste of green capacity at the neighboring leader intersection: in addition to minimizing its local queues, it also takes into account the expected queues that their switching decisions imply for their leader. We show, in realistic simulations, how the coordination introduced by this element allows to reduce queues not only at the leaders, but also at the follower nodes. The potential and key aspects of this scalable and simple coordinated control method for other applications is briefly discussed.