Nonlinear dynamics modelling for controller evolution

Julian Togelius, Renzo De Nardi, Hugo Marques, Richard Newcombe, Simon M. Lucas, Owen Holland

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

    Abstract

    The problem of how to acquire a model of a physical robot,which is fit for evolution of controllers that can subsequently be used to control that robot, is considered in the context of racing a radio-controlled toy car around a randomised track. Several modelling techniques are compared, and the specific properties of the acquired models that influence the quality of the evolved controller are discussed. As we aim tominimise the amount of domain knowledge used, we furtherinvestigate the relation between the assumptions about the modelled system made by particular modelling techniques and the suitability of the acquired models as bases for controller evolution. We find that none of the models acquired is good enough on its own, and that a key to evolving robustbehaviour is to evaluate controllers simultaneously on multiple models during evolution. Examples of successfully evolved racing control for the physical car are analysed.

    Original languageEnglish (US)
    Title of host publicationProceedings of GECCO 2007: Genetic and Evolutionary Computation Conference
    Pages324-333
    Number of pages10
    DOIs
    StatePublished - 2007
    Event9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007 - London, United Kingdom
    Duration: Jul 7 2007Jul 11 2007

    Other

    Other9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007
    CountryUnited Kingdom
    CityLondon
    Period7/7/077/11/07

    Fingerprint

    Nonlinear Modeling
    Dynamic Modeling
    Nonlinear Dynamics
    Controller
    Controllers
    Railroad cars
    Robot Control
    Robots
    Multiple Models
    Domain Knowledge
    Modeling
    Model
    Railroad tracks
    Robot
    Evaluate

    Keywords

    • Evolutionary robotics
    • Forward models
    • Games
    • Neural networks
    • System identification

    ASJC Scopus subject areas

    • Artificial Intelligence
    • Software
    • Theoretical Computer Science

    Cite this

    Togelius, J., De Nardi, R., Marques, H., Newcombe, R., Lucas, S. M., & Holland, O. (2007). Nonlinear dynamics modelling for controller evolution. In Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference (pp. 324-333) https://doi.org/10.1145/1276958.1277020

    Nonlinear dynamics modelling for controller evolution. / Togelius, Julian; De Nardi, Renzo; Marques, Hugo; Newcombe, Richard; Lucas, Simon M.; Holland, Owen.

    Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference. 2007. p. 324-333.

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

    Togelius, J, De Nardi, R, Marques, H, Newcombe, R, Lucas, SM & Holland, O 2007, Nonlinear dynamics modelling for controller evolution. in Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference. pp. 324-333, 9th Annual Genetic and Evolutionary Computation Conference, GECCO 2007, London, United Kingdom, 7/7/07. https://doi.org/10.1145/1276958.1277020
    Togelius J, De Nardi R, Marques H, Newcombe R, Lucas SM, Holland O. Nonlinear dynamics modelling for controller evolution. In Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference. 2007. p. 324-333 https://doi.org/10.1145/1276958.1277020
    Togelius, Julian ; De Nardi, Renzo ; Marques, Hugo ; Newcombe, Richard ; Lucas, Simon M. ; Holland, Owen. / Nonlinear dynamics modelling for controller evolution. Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference. 2007. pp. 324-333
    @inproceedings{775f4961616240c7be675e190cbe0e8a,
    title = "Nonlinear dynamics modelling for controller evolution",
    abstract = "The problem of how to acquire a model of a physical robot,which is fit for evolution of controllers that can subsequently be used to control that robot, is considered in the context of racing a radio-controlled toy car around a randomised track. Several modelling techniques are compared, and the specific properties of the acquired models that influence the quality of the evolved controller are discussed. As we aim tominimise the amount of domain knowledge used, we furtherinvestigate the relation between the assumptions about the modelled system made by particular modelling techniques and the suitability of the acquired models as bases for controller evolution. We find that none of the models acquired is good enough on its own, and that a key to evolving robustbehaviour is to evaluate controllers simultaneously on multiple models during evolution. Examples of successfully evolved racing control for the physical car are analysed.",
    keywords = "Evolutionary robotics, Forward models, Games, Neural networks, System identification",
    author = "Julian Togelius and {De Nardi}, Renzo and Hugo Marques and Richard Newcombe and Lucas, {Simon M.} and Owen Holland",
    year = "2007",
    doi = "10.1145/1276958.1277020",
    language = "English (US)",
    isbn = "1595936971",
    pages = "324--333",
    booktitle = "Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference",

    }

    TY - GEN

    T1 - Nonlinear dynamics modelling for controller evolution

    AU - Togelius, Julian

    AU - De Nardi, Renzo

    AU - Marques, Hugo

    AU - Newcombe, Richard

    AU - Lucas, Simon M.

    AU - Holland, Owen

    PY - 2007

    Y1 - 2007

    N2 - The problem of how to acquire a model of a physical robot,which is fit for evolution of controllers that can subsequently be used to control that robot, is considered in the context of racing a radio-controlled toy car around a randomised track. Several modelling techniques are compared, and the specific properties of the acquired models that influence the quality of the evolved controller are discussed. As we aim tominimise the amount of domain knowledge used, we furtherinvestigate the relation between the assumptions about the modelled system made by particular modelling techniques and the suitability of the acquired models as bases for controller evolution. We find that none of the models acquired is good enough on its own, and that a key to evolving robustbehaviour is to evaluate controllers simultaneously on multiple models during evolution. Examples of successfully evolved racing control for the physical car are analysed.

    AB - The problem of how to acquire a model of a physical robot,which is fit for evolution of controllers that can subsequently be used to control that robot, is considered in the context of racing a radio-controlled toy car around a randomised track. Several modelling techniques are compared, and the specific properties of the acquired models that influence the quality of the evolved controller are discussed. As we aim tominimise the amount of domain knowledge used, we furtherinvestigate the relation between the assumptions about the modelled system made by particular modelling techniques and the suitability of the acquired models as bases for controller evolution. We find that none of the models acquired is good enough on its own, and that a key to evolving robustbehaviour is to evaluate controllers simultaneously on multiple models during evolution. Examples of successfully evolved racing control for the physical car are analysed.

    KW - Evolutionary robotics

    KW - Forward models

    KW - Games

    KW - Neural networks

    KW - System identification

    UR - http://www.scopus.com/inward/record.url?scp=34548082582&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=34548082582&partnerID=8YFLogxK

    U2 - 10.1145/1276958.1277020

    DO - 10.1145/1276958.1277020

    M3 - Conference contribution

    SN - 1595936971

    SN - 9781595936974

    SP - 324

    EP - 333

    BT - Proceedings of GECCO 2007: Genetic and Evolutionary Computation Conference

    ER -