Design of an arterial signal progression plan for multi-path flows with only intersection turning counts

Abstract

Most congested urban arterials, connecting multiple major intersections with heavy turning flows, are likely to comprise multiple high-volume path flows in addition to its through traffic. Hence, such arterials, if designed with state-of-the-art models for progressing mainly through traffic flows, cannot achieve the expected level of performance, because those heavy path flows involving turning movements will inevitably suffer from overflows at their turning bays and consequently trigger a mutual blockage between the turning and through traffic flows. Although some recent advances on multi-path signal progression (Yang et al., 2015; Arsava et al., 2016) offer a viable design alternative for arterials accommodating such traffic patterns, their required data such as the arterial’s O-Ds or path volume information cannot be collected with state-of-the-practice traffic sensors, or mathematically estimated at the accuracy level sufficient for design of control strategies. As such, this study proposes a model with the optimized phase sequence and offsets to produce a progression band for each candidate path along the target arterial with only each intersection's volume counts, based on the set of “local bands” for all local paths constituted by each link’s upstream flow-in and downstream flow-out movements. Such identified local progression bands are then optimally connected between two neighboring links to form the progression band for traffic flows on each path to progress over multiple arterial links under the objective of maximizing the total weighted progression bands. To ensure the effectiveness of the produced progression bands, the proposed model has accounted for all factors and geometric constraints that may cause their mutual impedance in design of the offsets and phase sequences. The results of extensive numerical experiments with respect to delay and the number of stops per vehicle (reduced by 5.8% and 9.9%, respectively, compared to the benchmark model) within the entire network have shown the effectiveness of the proposed model.