Reinforcement learning approach to develop variable speed limit strategy using vehicle data and simulations
- Authors
- Kim, Yunjong; Kang, Kawon; Park, Nuri; Park, Juneyoung; Oh, Cheol
- Issue Date
- Jan-2024
- Publisher
- Taylor and Francis Ltd.
- Keywords
- Crash risk; proactive traffic safety management; reinforcement learning; variable speed limit; vehicle trajectory data
- Citation
- Journal of Intelligent Transportation Systems: Technology, Planning, and Operations, pp 1 - 18
- Pages
- 18
- Indexed
- SCIE
SSCI
SCOPUS
- Journal Title
- Journal of Intelligent Transportation Systems: Technology, Planning, and Operations
- Start Page
- 1
- End Page
- 18
- URI
- https://scholarworks.bwise.kr/erica/handle/2021.sw.erica/118829
- DOI
- 10.1080/15472450.2024.2312808
- ISSN
- 1547-2450
1547-2442
- Abstract
- A variety of studies have been conducted to evaluate real-time crash risk using vehicle trajectory data and to establish active traffic safety management measures. Speed management is an effective way to control traffic flow on freeways and to enhance safety. Currently, the variable speed limit (VSL) system is mainly applied in a limited manner during traffic congestion or bad weather. However, it is necessary to manage traffic safety proactively to prevent crashes by providing an appropriate target safety speed to minimize the real-time crash risk. Herein, a methodology for proactive traffic safety management is developed through speed management based on the estimation of real-time crash risk. The developed methodology evaluates performance through simulations and it consists of two components. First, a crash risk analyzer evaluates freeway crash risk by developing a real-time crash risk model based on real-world vehicle trajectory data matched with crash traffic flow. Then a speed manager implements a reinforcement learning-based VSL system in the simulation environment, which includes the crash risk derived from the crash risk analyzer through VISSIM-COM interfaces. The performance of the developed methodology was evaluated through VISSIM simulation analysis, and the results demonstrated its feasibility. The real-time crash risk was reduced by approximately 55% when the target safety speed information derived from the reinforcement learning model was provided in a scenario where one lane was closed due to a crash. The findings were further applied to establish an operations strategy for VSL systems based on both crash risk and actual traffic conditions. © 2024 Taylor & Francis Group, LLC.
- Files in This Item
-
Go to Link
- Appears in
Collections - COLLEGE OF ENGINEERING SCIENCES > DEPARTMENT OF TRANSPORTATION AND LOGISTICS ENGINEERING > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/erica/handle/2021.sw.erica/118829)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.