CoLight: Learning Network-level Cooperation for Traffic Signal Control
This addresses traffic congestion in urban environments by improving signal coordination, though it is an incremental advancement in applying graph networks to reinforcement learning for traffic control.
The paper tackled the problem of dynamic traffic signal control by proposing CoLight, a model that uses graph attentional networks to enable cooperation among intersections, achieving superior performance against state-of-the-art methods in experiments on large-scale networks with hundreds of signals.
Cooperation among the traffic signals enables vehicles to move through intersections more quickly. Conventional transportation approaches implement cooperation by pre-calculating the offsets between two intersections. Such pre-calculated offsets are not suitable for dynamic traffic environments. To enable cooperation of traffic signals, in this paper, we propose a model, CoLight, which uses graph attentional networks to facilitate communication. Specifically, for a target intersection in a network, CoLight can not only incorporate the temporal and spatial influences of neighboring intersections to the target intersection, but also build up index-free modeling of neighboring intersections. To the best of our knowledge, we are the first to use graph attentional networks in the setting of reinforcement learning for traffic signal control and to conduct experiments on the large-scale road network with hundreds of traffic signals. In experiments, we demonstrate that by learning the communication, the proposed model can achieve superior performance against the state-of-the-art methods.