A Clustering Approach for Real-Time Control of Queue Spillbacks on Signalized Arterials

When demand exceeds capacity at signalized intersections, queues fail to clear during the allocated green times.  This creates oversaturated traffic conditions and when queues reach the upstream intersection a queue spillback occurs that reduces the upstream link capacity in arterials. To mitigate the negative impacts of spillbacks, this paper presents a traffic signal methodology for global management of long queues along signalized arterials. The key idea of the proposed methodology is to implement a real-time partitioning of the arterial to detect critical group(s) of consecutive links with oversaturated traffic conditions. The clustering approach enables to develop locally smaller-sized decentralized signal control strategies operating on the most upstream and downstream intersections of the critical group, with respect to the empty space available in the congested links beyond the back of queue. Simulation investigations on a real-world arterial site demonstrate the benefits of this new approach compared to an existing pre-timed signal control strategy and a classical decentralized green extension strategy.