Abstract:
This paper advances the state-of-the-art of traffic signal control in congested urban arterials by integrating a clustering approach into a control policy of long queues prior to spillover occurrence to reduce the risk of spillovers through a feedback strategy. First, the authors introduce an arterial clustering approach that detects in real-time the links with long queues along one direction of the arterial, clustering them together if they are consecutive and then identifying the entrance and exit intersections of each cluster. These intersections indeed constitute critical junctions (considering the analogy of an active bottleneck in a freeway) and therefore, proper adjustment of the signal timing settings in those intersections could improve traffic conditions in the whole arterial. Thus, it enables to implement locally smaller-sized decentralized signal control strategies while ensuring at the same time the global coherence of these strategies along the arterial. In this manner, they seek to improve traffic on arterial at a very low cost by acting merely on critical intersections, as opposed for instance to network-wide optimization process. This approach is adaptive and does not require information about turning movements, which is difficult to be estimated in real-time. Hence, the purpose of this paper is to develop an elegant signal control strategy based on the arterial clustering approach that enables to act only locally on specific intersections. Including an advanced detection of oversaturated states and a specific focus on queue spillovers prevention, it leads to significant reduction of congestion and thus improves the network traffic conditions.
