Reinforcement learning offers a novel approach for addressing complicated signal optimization challenges at intersections. Previous research was conducted to conventional traffic detection methods, resulting in imprecise traffic information. However, with technological advancements, access to highly accurate traffic data becomes possible through advanced detector technology. This serves as a reliable data source for deep reinforcement learning. Given the numerous intersections in urban networks, effectively applying reinforcement learning in real-world situations requires addressing the challenge of transferring knowledge gains during training of new scenarios. This study presents a novel approach to traffic control system design using reinforcement learning, focusing on evaluating its performance in terms of waiting time and standard deviation. Compared to existing models, the proposed system demonstrates remarkable improvements, with significantly reduced waiting times and standard deviation. The low waiting time highlights the system’s efficiency in managing traffic flow, leading to reduced congestion and a better commuting experience. Moreover, the decreased standard deviation signifies a more predictable traffic pattern, contributing to safety and reliability in travel times. These findings hold promise for improving urban mobility and reducing the environmental impact of traffic congestion. By harnessing reinforcement learning, cities can aspire to implement more sustainable transportation systems, benefiting both commuters and the environment. The results indicate that the presented approach offered least waiting time up to 0 for emergency vehicles, 40 seconds for regular vehicles and lower standard deviation (SD) in contrast to the existing methods.
Reinforcement Learning Model for Traffic Control in Urban Area
24.11.2023
500965 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Urban traffic control method based on reinforcement learning
Europäisches Patentamt | 2024
|Adaptive Traffic Signal Control for Urban Corridor Based on Reinforcement Learning
Springer Verlag | 2024
|