Development of Monte-Carlo-Optimization-Based Variable Speed Limit Using Microscopic Traffic Flow Simulation on Freeways

• Published in: Journal of the Eastern Asia Society for Transportation Studies Vol. 15; pp. 2804 - 2817
• Main Authors: FUJIMOTO, So, SHIOMI, Yasuhiro, HANABUSA, Hisatomo
• Format: Journal Article
• Language: English
• Published: Eastern Asia Society for Transportation Studies 2024
• Subjects: Freeway; Microscopic Simulation; Reinforcement Learning; Variable Speed Limit
• ISSN: 1881-1124
• DOI: 10.11175/easts.15.2804

Variable Speed Limit (VSL) is widely used to mitigate congestion on freeways. However, the current variable speed limit control is not always optimal and does not respond well to real traffic flows. Therefore, this study proposes a speed limit control method on freeways to reduce congestion and travel time on freeways by applying microscopic-simulation-based reinforcement learning model. The proposed speed control system consists of a reinforcement learning model and a traffic situation prediction model based on microscopic traffic flow simulation. By applying reinforcement learning, the system learns the optimal speed limit values under different traffic conditions and calculates the optimal speed limit for each road section. The validity of the constructed speed control system was verified and improvements in traffic conditions, such as a reduction in average travel time, were observed with and without the control.