Delay-throughput tradeoffs for signalized networks with finite queue capacity

Journal article, 2024

Network-level adaptive signal control is an effective way to reduce delay and increase network throughput. However, in the face of asymmetric exogenous demand, the increase of network performance via adaptive signal control alone is at the expense of service fairness (i.e., phase actuation fairness and network resource utilization fairness). In addition, for oversaturated networks, arbitrary adaptive signal control seems to have little effect on improving network performance. Therefore, under the assumption that the mean routing proportions/turn ratios of vehicles at intersections are fixed, this study investigates the problem of optimally allocating input rates to entry links and simultaneously finding a stabilizing signal control policy with phase fairness. We model the stochastic optimization problem of maximizing network throughput subject to network stability (i.e., all queue lengths have finite means) and average phase actuation constraints to bridge the gap between stochastic network stability control and convex optimization. Moreover, we further propose a micro-level joint admission and bounded signal control algorithm to achieve network stability and throughput optimization simultaneously. Joint control is implemented in a fully decomposed and distributed manner. For any arrival rate, joint control provably achieves network throughput within O(1/V) of optimality while trading off average delay with O(V), where V is an adjusted control parameter. Through a comparative simulation of a real network with 256 O-D pairs, the proposed joint control keeps network throughput at maximum, guarantees service fairness, and fully utilizes network capacity (i.e., increases network throughput by 17.54%).