We describe an approach to creating a controller for The Open Car Racing Simulator (TORCS), based on The Simulated Car Racing Championship (SCRC) client, using unsupervised evolutionary learning for recurrent neural networks. Our method of training the recurrent neural network controllers relies on combining the components of the singular value decomposition of two different neural network connection matrices.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Training RNN simulated vehicle controllers using the SVD and evolutionary algorithms


    Beteiligte:


    Erscheinungsdatum :

    01.06.2018


    Format / Umfang :

    675796 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    TRAINING RNN SIMULATED VEHICLE CONTROLLERS USING THE SVD AND EVOLUTIONARY ALGORITHMS

    McNeill, Daniel K. | British Library Conference Proceedings | 2018



    Optimizing traffic light controllers by means of evolutionary algorithms

    Taale, H. / Bäck, T. / Preuß, M. et al. | Tema Archiv | 1998


    Evolutionary Design of Robot Controllers

    Prof. Watanabe, Keigo / Prof. Hashem, M. M. A. | Springer Verlag | 2004