Robust Traffic Signal Retiming Based on Queue Service Time Estimation Using Low-Penetration Connected Vehicle Data

Signal retiming is the most cost-efficient measure to reduce vehicle delay and alleviate congestion on urban roads. Previous studies have explored the potential of using connected vehicle data for signal retiming specifically under the current low-penetration environment, which will significantly reduce the cost and increase the productivity of signal retiming. However, the existing methods are mostly deterministic and do not well consider the uncertainty in both traffic demand and capacity. This compromises their robustness in a real application. In this study, a novel traffic state measure—queue service time (QST)—is introduced and used as the only input to generate a robust signal plan at isolated intersections for a time-of-day period. First, a Bayesian-based model is proposed to estimate the QST distribution by collectively using the lower and upper boundary observations reported by connected vehicles. Then, a goal programming-based signal optimization model is formulated using quantiles of QST as input, which accounts for the combined uncertainty in both traffic demand and capacity. Simulation experiments validate the effectiveness and robustness of the proposed method. It is shown that the proposed QST estimation model is reliable to use under a penetration rate as low as 0.05 and can effectively estimate the actual distribution in both under- and oversaturation conditions. Compared with a demand-based method that only accounts for uncertainty in traffic demand, the proposed QST-based signal timing optimization method shows its superiority in reducing the occurrence of oversaturation or phase failure, as well as enhancing performance against the worst cases.