A COORDINATED SIGNAL CONTROL SYSTEM FOR COMMUTING ARTERIALS WITH HIGHLY ASYMMETRIC DIRECTIONAL FLOWS

Abstract

Urban commuting arterials often encounter unique operational challenges during peak periods due to highly asymmetric directional traffic patterns and the frequent queue spillbacks and mutual blockages. These traffic disruptions—particularly at intersections with short turning bays and limited link storage—can severely degrade the progression capacity and cause excessive delays, especially for high-volume turning and through movements. While considerable progress has been made by the traffic community over the years in developing various signal control systems, most existing strategies—both in traditional time-of-day coordination and advanced real-time adaptive control—have not fully addressed the complexities of mid-block congestion dynamics, cross-lane interference, and proactive mitigation of queue spillbacks as well as mutual blockages.This dissertation introduces a modular-based and scalable arterial signal control system specifically designed for congested commuting arterials with directional demand imbalances and short turning-bay lengths. The developed system comprises three integrated modules: an offline time-of-day control module, a responsive real-time control module, and a proactive real-time control module. The time-of-day module is designed to produce the optimized phase sequences and signal offsets by integrating delay minimization for the high-volume commuting direction with the progression operations for the opposite direction, while accounting for the roadway geometric constraints such as turning-bay and short link lengths. The responsive module functions to advance the pre-timed signal plans to real-time adaptation through a rule-based logic structure that detects and classifies congestion patterns, and then applies the selected signal adjustments to targeted intersections so as to mitigate the traffic impacts of queue spillbacks and lane blockages. Advancing the core control notion from responsive to proactive, this study further develops a proactive signal control module, which is incorporated with a multi-branch, multi-head Long Short-Term Memory (LSTM) architecture to predict queue states, based on the complex spatial-temporal dependencies of traffic states and queue patterns between the target intersection and its multiple upstream intersections. This embedded predictive function allows the system to execute preemptive control actions before the onset of critical congestion patterns. To ensure the operational reliability and stability, a real-time monitoring mechanism for evaluating the accuracy of prediction, triggering the fallback to responsive control when necessary, has also been included in this proactive signal module. The developed system has been evaluated through traffic simulation experiments using a real-world arterial testbed on MD 355 Rockville Pike in Bethesda, Maryland, under evening peak-hour conditions. Experimental results with respect to common measures of effectiveness have confirmed the promising properties of the developed system, where 1) its offline time-of-day module appears to outperform state-of-the-art signal coordination methods such as MULTIBAND and TRANSYT-7F under congested peak-hour traffic conditions; 2) the responsive real-time module has further improved the time-of-day control’s performance on dynamically mitigating queue spillbacks and mutual blockages; and 3) the system’s advanced proactive module can achieve the highest overall performance with its unique functions to predict the traffic state and congestion patterns in the predefined time horizon and then execute the optimal preemptive signal control actions in advance. Overall, the experimental results from extensive traffic scenarios have collectively confirmed the flexibility and effectiveness of the developed system under various information availability, control environments, and geometric constraints, as well as asymmetric volume patterns over congested commuting arterials. With the support of state-of-the-art traffic sensors, the developed system offers an effective alternative for use in practice to contend with ever-increasing traffic congestion in urban commuting corridors.