Improving Adaptive Traffic Signal Control at Urban Intersections Based on Artificial Intelligence

This article addresses the problem of improving an adaptive traffic signal control algorithm at urban intersections by taking into account the dynamic characteristics of traffic flow. Fixed-time control under conditions of rapidly changing traffic demand may lead to excessive waiting times and increased queue lengths. To address this problem, an adaptive control approach based on reinforcement learning principles is proposed. The approach considers the current state of traffic flow, queue lengths, waiting times, and service balance among different traffic directions. The article describes the algorithm's state vector, control actions, reward function, and safety constraints. To evaluate the proposed approach, an experimental methodology is developed using the SUMO microscopic traffic simulator and the TraCI interface.