TY - JOUR
AU - Nourmohammadi, Hamzehali
AB - Smart cities seek to promote specific strategies for addressing problems like traffic, parking, pollution, and security caused by the rapid urbanization and population growth. Hence, it is essential to control urban traffic using efficient traffic light scheduling tactics. This approach proposes fuzzy reinforcement learning based real-time traffic light control solution. The reward system is based on the number of cars waiting in line and the time it takes for the traffic light to turn green. Based on the traffic patterns at various times, the model dynamically determines the time to alter the illumination phase of the light. A fuzzy logic system trained using a genetic or wild horse optimization algorithm will assist in making this judgment. Two alternative scenarios are provided for the simulations: static (where the timing is fixed at 60 s for the green signal), and dynamic (where the timing varies in real time depending on the current congestion level and trained fuzzy decision making). A few key evaluation parameters, including waiting time and traffic density, were applied to demonstrate the efficacy of the suggested approach. The fuzzy rules in the reinforcement learning system based on fuzzy logic are determined by these two criteria as defined rewards. The training by the WHO algorithm produced the optimal timing, based on simulation results on a traffic signal in a case study intersection during the day and night. In comparison to the fixed scheduling strategy, it has also been able to minimize traffic congestion by 52% and queue delays by 67%.
TI - Optimizing and Managing the Lighting Time of the Traffic Light Using the Reinforcement Learning System Based on Fuzzy Logic and Training the System with Evolutionary Algorithms
JF - International Journal of Intelligent Transportation Systems Research
DO - 10.1007/s13177-025-00504-w
DA - 2025-06-04
UR - https://www.deepdyve.com/lp/springer-journals/optimizing-and-managing-the-lighting-time-of-the-traffic-light-using-YmHRzJk41q
SP - 1
EP - 20
VL - OnlineFirst
IS - 
DP - DeepDyve
ER -