Adaptive Traffic Signal Control via Stackelberg Game Theory and Q-Learning

This paper proposes a hybrid intelligent traffic signal control framework combining Stackelberg game theory and Q-Learning to optimize intersection management. The intersection is modeled as a non-exhaustive M/M/1 polling system with Binomial Gated Service and multiple vacations, capturing stochastic traffic behavior and congestion priorities. The Stackelberg game determines optimal green-light durations through hierarchical decision-making, while Q- Learning enables real-time adaptation to changing traffic conditions. Simulation results in SUMO show significant reductions in waiting times and queue lengths, along with improved overall traffic flow efficiency.