Data-driven multi-agent vehicle routing in a congested city

Alex Solter, Fuhua Lin, Dunwei Wen, Xiaokang Zhou

    Research output: Contribution to journalJournal Articlepeer-review

    Abstract

    Navigation in a traffic congested city can prove to be a difficult task. Often a path that may appear to be the fastest option is much slower due to congestion. If we can predict the effects of congestion, it may be possible to develop a better route that allows us to reach our destination more quickly. This paper studies the possibility of using a centralized real-time traffic information system containing travel time data collected from each road user. These data are made available to all users, such that they may be able to learn and predict the effects of congestion for building a route adaptively. This method is further enhanced by combining the traffic information system data with previous routing experiences to determine the fastest route with less exploration. We test our method using a multi-agent simulation, demonstrating that this method produces a lower total route time for all vehicles than when using either a centralized traffic information system or direct experience alone.

    Original languageEnglish
    Article number447
    JournalInformation (Switzerland)
    Volume12
    Issue number11
    DOIs
    Publication statusPublished - Nov. 2021

    Keywords

    • Multi-agent simulation
    • Multi-armed bandit
    • Traffic information system
    • Vehicle routing

    Fingerprint

    Dive into the research topics of 'Data-driven multi-agent vehicle routing in a congested city'. Together they form a unique fingerprint.

    Cite this