Advanced traffic signal control system using deep double q-learning with pedestrian factors

In response to the increasingly severe traffic congestion problem, this paper proposes a novel method based on Double Deep Q-Learning Network to enhance the performance of adaptive traffic signal control agents in alleviating traffic congestion and delays. By designing a novel state space model and reward function, the proposed method can minimize vehicle queue lengths and reduce vehicle delay duration when dealing with complex intersections or segments with significant traffic fluctuations. To evaluate the performance of this method, the paper utilizes the Simulation of Urban MObility software to set up environments for complex intersections. Simulation results demonstrate that compared to previous works and current mainstream algorithms, the proposed method can efficiently control signals in complex traffic environments, effectively addressing congestion and improving traffic efficiency.

---

Opracowanie rekordu ze środków MNiSW, umowa nr POPUL/SP/0154/2024/02 w ramach programu "Społeczna odpowiedzialność nauki II" - moduł: Popularyzacja nauki (2025).