Racing game artificial intelligence using evolutionary artificial neural networks

Hamid Homatash, Süheyl Özveren, Victor Bassilious

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


    This paper investigates the viability of using an Evolutionary Artificial Neural Network (EANN) approach as an alternative to standard Artificial Intelligence techniques used in a racing game. Use of Neuro-Evolution of Augmenting Topologies (NEAT) algorithms is compared to a standard AI technique which employs steering behaviours and a finite state machine to navigate an AI- driver agent around a circuit. We present a comparison between the NEAT algorithm and the standard AI technique described. Our initial literature review of the different available EANN approaches and the reasons for the choice of the NEAT algorithmic approach for our investigations is followed by the description of the implementation of our modified NEAT algorithm based EANN AI-driver agent and the racing simulation used for testing. Finally comparison of the results achieved with the implemented NEAT algorithm and the standard AI technique is followed by our conclusion on the comparative effectiveness of the NEAT and standard AI-driver agents and how our reported results can be further improved by future studies
    Original languageEnglish
    Title of host publicationGame-On AI 2011
    Number of pages11
    Publication statusPublished - 22 Aug 2011


    • machine learning
    • games
    • neural networks
    • artificial intelligence
    • EANN
    • racing game simulator


    Dive into the research topics of 'Racing game artificial intelligence using evolutionary artificial neural networks'. Together they form a unique fingerprint.

    Cite this