Optimizing Traffic Flow with Policy Iteration Method: A Study on 9-Intersection Network Management

The foundation of urban traffic management lies in the effective control of intersections with traffic lights. Despite this, the lack of adaptive management strategies in response to changes in traffic patterns results in significant and frustrating delays for motorists. This paper proposes the use of a Markov Decision Process (MDP) to decrease waiting times at intersections. The MDP considers the network environment, encompassing limitations related to both its configuration and protection.Several studies have tackled this issue by examining each intersection in isolation, disregarding its interactions with other intersections. In contrast, the current study takes a systemic approach to the network of intersections. It adopts a non linear approach that takes into account the interconnection between interjections. The highest number of intersections managed in previous studies is eight while it is nine in the model proposed in this study. Given that the Policy Iteration (PI) method is well-suited to the proposed MDP, it has been selected as the resolution method. The PI method, compared to systems with fixed phase durations, has effectively reduced the average waiting time up to 30% compared to non-adaptive systems.