Online learning of humanoid robot kinematics under switching tools contexts

Lorenzo Jamone, Bruno Damas, Jose Santos-Victor, Atsuo Takanishi

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

    7 Citations (Scopus)

    Abstract

    In this paper a novel approach to kinematics learning and task space control, under switching contexts, is presented. Such non-stationary contexts may appear in many robotic tasks: in particular, the changing of the context due to the use of tools with different lengths and shapes is herein studied. We model the robot forward kinematics as a multi-valued function, in which different outputs for the same input query are related to actual different hidden contexts. To do that, we employ IMLE, a recent online learning algorithm that fits an infinite mixture of linear experts to the online stream of training data. This algorithm can directly provide multi-valued regression in a online fashion, while having, for classic single-valued regression, a performance comparable to state-of-the-art online learning algorithms. The context varying forward kinematics is learned online through exploration, not relying on any kind of prior knowledge. Using the proposed approach, the robot can dynamically learn how to use different tools, without forgetting the kinematic mappings concerning previously manipulated tools. No information is given about such tool changes to the learning algorithm, nor any assumption is made about the tool kinematics. To our knowledge this is the most general and efficient approach to learning and control under discrete varying contexts. Some experimental results obtained on a high-dimensional simulated humanoid robot provide a strong support to our approach.

    Original languageEnglish
    Title of host publicationProceedings - IEEE International Conference on Robotics and Automation
    Pages4811-4817
    Number of pages7
    DOIs
    Publication statusPublished - 2013
    Event2013 IEEE International Conference on Robotics and Automation, ICRA 2013 - Karlsruhe
    Duration: 2013 May 62013 May 10

    Other

    Other2013 IEEE International Conference on Robotics and Automation, ICRA 2013
    CityKarlsruhe
    Period13/5/613/5/10

    Fingerprint

    Kinematics
    Robots
    Learning algorithms
    Robotics

    ASJC Scopus subject areas

    • Software
    • Artificial Intelligence
    • Control and Systems Engineering
    • Electrical and Electronic Engineering

    Cite this

    Jamone, L., Damas, B., Santos-Victor, J., & Takanishi, A. (2013). Online learning of humanoid robot kinematics under switching tools contexts. In Proceedings - IEEE International Conference on Robotics and Automation (pp. 4811-4817). [6631263] https://doi.org/10.1109/ICRA.2013.6631263

    Online learning of humanoid robot kinematics under switching tools contexts. / Jamone, Lorenzo; Damas, Bruno; Santos-Victor, Jose; Takanishi, Atsuo.

    Proceedings - IEEE International Conference on Robotics and Automation. 2013. p. 4811-4817 6631263.

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

    Jamone, L, Damas, B, Santos-Victor, J & Takanishi, A 2013, Online learning of humanoid robot kinematics under switching tools contexts. in Proceedings - IEEE International Conference on Robotics and Automation., 6631263, pp. 4811-4817, 2013 IEEE International Conference on Robotics and Automation, ICRA 2013, Karlsruhe, 13/5/6. https://doi.org/10.1109/ICRA.2013.6631263
    Jamone L, Damas B, Santos-Victor J, Takanishi A. Online learning of humanoid robot kinematics under switching tools contexts. In Proceedings - IEEE International Conference on Robotics and Automation. 2013. p. 4811-4817. 6631263 https://doi.org/10.1109/ICRA.2013.6631263
    Jamone, Lorenzo ; Damas, Bruno ; Santos-Victor, Jose ; Takanishi, Atsuo. / Online learning of humanoid robot kinematics under switching tools contexts. Proceedings - IEEE International Conference on Robotics and Automation. 2013. pp. 4811-4817
    @inproceedings{786275427cb84d2385aa11d54f84a026,
    title = "Online learning of humanoid robot kinematics under switching tools contexts",
    abstract = "In this paper a novel approach to kinematics learning and task space control, under switching contexts, is presented. Such non-stationary contexts may appear in many robotic tasks: in particular, the changing of the context due to the use of tools with different lengths and shapes is herein studied. We model the robot forward kinematics as a multi-valued function, in which different outputs for the same input query are related to actual different hidden contexts. To do that, we employ IMLE, a recent online learning algorithm that fits an infinite mixture of linear experts to the online stream of training data. This algorithm can directly provide multi-valued regression in a online fashion, while having, for classic single-valued regression, a performance comparable to state-of-the-art online learning algorithms. The context varying forward kinematics is learned online through exploration, not relying on any kind of prior knowledge. Using the proposed approach, the robot can dynamically learn how to use different tools, without forgetting the kinematic mappings concerning previously manipulated tools. No information is given about such tool changes to the learning algorithm, nor any assumption is made about the tool kinematics. To our knowledge this is the most general and efficient approach to learning and control under discrete varying contexts. Some experimental results obtained on a high-dimensional simulated humanoid robot provide a strong support to our approach.",
    author = "Lorenzo Jamone and Bruno Damas and Jose Santos-Victor and Atsuo Takanishi",
    year = "2013",
    doi = "10.1109/ICRA.2013.6631263",
    language = "English",
    isbn = "9781467356411",
    pages = "4811--4817",
    booktitle = "Proceedings - IEEE International Conference on Robotics and Automation",

    }

    TY - GEN

    T1 - Online learning of humanoid robot kinematics under switching tools contexts

    AU - Jamone, Lorenzo

    AU - Damas, Bruno

    AU - Santos-Victor, Jose

    AU - Takanishi, Atsuo

    PY - 2013

    Y1 - 2013

    N2 - In this paper a novel approach to kinematics learning and task space control, under switching contexts, is presented. Such non-stationary contexts may appear in many robotic tasks: in particular, the changing of the context due to the use of tools with different lengths and shapes is herein studied. We model the robot forward kinematics as a multi-valued function, in which different outputs for the same input query are related to actual different hidden contexts. To do that, we employ IMLE, a recent online learning algorithm that fits an infinite mixture of linear experts to the online stream of training data. This algorithm can directly provide multi-valued regression in a online fashion, while having, for classic single-valued regression, a performance comparable to state-of-the-art online learning algorithms. The context varying forward kinematics is learned online through exploration, not relying on any kind of prior knowledge. Using the proposed approach, the robot can dynamically learn how to use different tools, without forgetting the kinematic mappings concerning previously manipulated tools. No information is given about such tool changes to the learning algorithm, nor any assumption is made about the tool kinematics. To our knowledge this is the most general and efficient approach to learning and control under discrete varying contexts. Some experimental results obtained on a high-dimensional simulated humanoid robot provide a strong support to our approach.

    AB - In this paper a novel approach to kinematics learning and task space control, under switching contexts, is presented. Such non-stationary contexts may appear in many robotic tasks: in particular, the changing of the context due to the use of tools with different lengths and shapes is herein studied. We model the robot forward kinematics as a multi-valued function, in which different outputs for the same input query are related to actual different hidden contexts. To do that, we employ IMLE, a recent online learning algorithm that fits an infinite mixture of linear experts to the online stream of training data. This algorithm can directly provide multi-valued regression in a online fashion, while having, for classic single-valued regression, a performance comparable to state-of-the-art online learning algorithms. The context varying forward kinematics is learned online through exploration, not relying on any kind of prior knowledge. Using the proposed approach, the robot can dynamically learn how to use different tools, without forgetting the kinematic mappings concerning previously manipulated tools. No information is given about such tool changes to the learning algorithm, nor any assumption is made about the tool kinematics. To our knowledge this is the most general and efficient approach to learning and control under discrete varying contexts. Some experimental results obtained on a high-dimensional simulated humanoid robot provide a strong support to our approach.

    UR - http://www.scopus.com/inward/record.url?scp=84887293509&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84887293509&partnerID=8YFLogxK

    U2 - 10.1109/ICRA.2013.6631263

    DO - 10.1109/ICRA.2013.6631263

    M3 - Conference contribution

    SN - 9781467356411

    SP - 4811

    EP - 4817

    BT - Proceedings - IEEE International Conference on Robotics and Automation

    ER -