Point-to-point car racing

An initial study of evolution versus temporal difference learning

Simon M. Lucas, Julian Togelius

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    This paper considers variations on an extremely simple form of car racing, the challenge being to visit as many way-points as possible in a fixed amount of time. The simplicity of the models enables a very thorough evaluation of various learning algorithms and control architectures, and enables other researchers to work on the same models with relative ease. The models are used to compare the performance of various hand-programmed controllers, and neural networks trained using evolution, and using temporal difference learning. Comparisons are also made between state-based and action-based controller architectures. The best controllers were obtained using evolution to learn the weights of state-evaluation neural networks, and these were greatly superior to human drivers.

    Original languageEnglish (US)
    Title of host publicationProceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007
    Pages260-267
    Number of pages8
    DOIs
    StatePublished - 2007
    Event2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007 - Honolulu, HI, United States
    Duration: Apr 1 2007Apr 5 2007

    Other

    Other2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007
    CountryUnited States
    CityHonolulu, HI
    Period4/1/074/5/07

    Fingerprint

    Railroad cars
    Controller
    Controllers
    Neural Networks
    Neural networks
    Learning Control
    Evaluation
    Learning algorithms
    Driver
    Learning Algorithm
    Simplicity
    Model
    Learning
    Architecture
    Human
    Form

    Keywords

    • Car racing
    • Evolving neural networks
    • Reinforcement learning

    ASJC Scopus subject areas

    • Artificial Intelligence
    • Electrical and Electronic Engineering
    • Computational Mathematics
    • Theoretical Computer Science

    Cite this

    Lucas, S. M., & Togelius, J. (2007). Point-to-point car racing: An initial study of evolution versus temporal difference learning. In Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007 (pp. 260-267). [4219052] https://doi.org/10.1109/CIG.2007.368107

    Point-to-point car racing : An initial study of evolution versus temporal difference learning. / Lucas, Simon M.; Togelius, Julian.

    Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007. 2007. p. 260-267 4219052.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Lucas, SM & Togelius, J 2007, Point-to-point car racing: An initial study of evolution versus temporal difference learning. in Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007., 4219052, pp. 260-267, 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007, Honolulu, HI, United States, 4/1/07. https://doi.org/10.1109/CIG.2007.368107
    Lucas SM, Togelius J. Point-to-point car racing: An initial study of evolution versus temporal difference learning. In Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007. 2007. p. 260-267. 4219052 https://doi.org/10.1109/CIG.2007.368107
    Lucas, Simon M. ; Togelius, Julian. / Point-to-point car racing : An initial study of evolution versus temporal difference learning. Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007. 2007. pp. 260-267
    @inproceedings{db2d53168fc6422e8a263687041a891c,
    title = "Point-to-point car racing: An initial study of evolution versus temporal difference learning",
    abstract = "This paper considers variations on an extremely simple form of car racing, the challenge being to visit as many way-points as possible in a fixed amount of time. The simplicity of the models enables a very thorough evaluation of various learning algorithms and control architectures, and enables other researchers to work on the same models with relative ease. The models are used to compare the performance of various hand-programmed controllers, and neural networks trained using evolution, and using temporal difference learning. Comparisons are also made between state-based and action-based controller architectures. The best controllers were obtained using evolution to learn the weights of state-evaluation neural networks, and these were greatly superior to human drivers.",
    keywords = "Car racing, Evolving neural networks, Reinforcement learning",
    author = "Lucas, {Simon M.} and Julian Togelius",
    year = "2007",
    doi = "10.1109/CIG.2007.368107",
    language = "English (US)",
    isbn = "1424407095",
    pages = "260--267",
    booktitle = "Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007",

    }

    TY - GEN

    T1 - Point-to-point car racing

    T2 - An initial study of evolution versus temporal difference learning

    AU - Lucas, Simon M.

    AU - Togelius, Julian

    PY - 2007

    Y1 - 2007

    N2 - This paper considers variations on an extremely simple form of car racing, the challenge being to visit as many way-points as possible in a fixed amount of time. The simplicity of the models enables a very thorough evaluation of various learning algorithms and control architectures, and enables other researchers to work on the same models with relative ease. The models are used to compare the performance of various hand-programmed controllers, and neural networks trained using evolution, and using temporal difference learning. Comparisons are also made between state-based and action-based controller architectures. The best controllers were obtained using evolution to learn the weights of state-evaluation neural networks, and these were greatly superior to human drivers.

    AB - This paper considers variations on an extremely simple form of car racing, the challenge being to visit as many way-points as possible in a fixed amount of time. The simplicity of the models enables a very thorough evaluation of various learning algorithms and control architectures, and enables other researchers to work on the same models with relative ease. The models are used to compare the performance of various hand-programmed controllers, and neural networks trained using evolution, and using temporal difference learning. Comparisons are also made between state-based and action-based controller architectures. The best controllers were obtained using evolution to learn the weights of state-evaluation neural networks, and these were greatly superior to human drivers.

    KW - Car racing

    KW - Evolving neural networks

    KW - Reinforcement learning

    UR - http://www.scopus.com/inward/record.url?scp=34548765403&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=34548765403&partnerID=8YFLogxK

    U2 - 10.1109/CIG.2007.368107

    DO - 10.1109/CIG.2007.368107

    M3 - Conference contribution

    SN - 1424407095

    SN - 9781424407095

    SP - 260

    EP - 267

    BT - Proceedings of the 2007 IEEE Symposium on Computational Intelligence and Games, CIG 2007

    ER -