Towards intrinsically learned skin models in robots

Daniel Polani, Simon Mcgregor

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

    1 Citation (Scopus)

    Abstract

    We describe the use of information-theoretic principles following [4] for sensoritopic map construction on a simulated "skin". This research constitutes a first step in robot body model learning based on uninterpreted tactile sensors. Results are reasonable, although our findings identify a serious limitation in the information distance metric: the distance is analytically bounded by the number of bins used for discretisation, so that reconstructions tend to "spherise". Furthermore, the tactile modality is significantly sparser than the visual modality, which introduces further artefacts in the information distance, particularly under maximum entropy binning. We conclude that some method which allows sensoritopic reconstruction directly from mutual information would be preferable.
    Original languageEnglish
    Title of host publicationProceedings of the 2nd International Symposium on New Frontiers in Human-Robot Interaction - A Symposium at the AISB 2010 Convention
    Pages72-75
    Number of pages4
    Publication statusPublished - 1 Dec 2010
    Event2nd International Symposium on New Frontiers in Human-Robot Interaction - A Symposium at the AISB 2010 Convention - Leicester, United Kingdom
    Duration: 29 Mar 20101 Apr 2010

    Publication series

    NameProceedings of the 2nd International Symposium on New Frontiers in Human-Robot Interaction - A Symposium at the AISB 2010 Convention

    Conference

    Conference2nd International Symposium on New Frontiers in Human-Robot Interaction - A Symposium at the AISB 2010 Convention
    Country/TerritoryUnited Kingdom
    CityLeicester
    Period29/03/101/04/10

    Fingerprint

    Dive into the research topics of 'Towards intrinsically learned skin models in robots'. Together they form a unique fingerprint.

    Cite this