Learning task space control through goal directed exploration

Lorenzo Jamone, Lorenzo Natale, Kenji Hashimoto, Giulio Sandini, Atsuo Takanishi

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

    17 Citations (Scopus)

    Abstract

    We present an autonomous goal-directed strategy to learn how to control a redundant robot in the task space. We discuss the advantages of exploring the state space through goal-directed actions defined in the task space (i.e. learning by trying to do) instead of performing motor babbling in the joints space, and we stress the importance of learning to be performed online, without any separation between training and execution. Our solution relies on learning the forward model and then inverting it for the control; different approaches to learn the forward model are described and compared. Experimental results on a simulated humanoid robot are provided to support our claims. The robot learns autonomously how to perform reaching actions directed toward 3D targets in task space by using arm and waist motion, not relying on any prior knowledge or initial motor babbling. To test the ability of the system to adapt to sudden changes both in the robot structure and in the perceived environment we artificially introduce two different kinds of kinematic perturbations: a modification of the length of one link and a rotation of the task space reference frame. Results demonstrate that the online update of the model allows the robot to cope with such situations.

    Original languageEnglish
    Title of host publication2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011
    Pages702-708
    Number of pages7
    DOIs
    Publication statusPublished - 2011
    Event2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011 - Phuket
    Duration: 2011 Dec 72011 Dec 11

    Other

    Other2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011
    CityPhuket
    Period11/12/711/12/11

    Fingerprint

    Robots
    Kinematics

    ASJC Scopus subject areas

    • Computer Vision and Pattern Recognition

    Cite this

    Jamone, L., Natale, L., Hashimoto, K., Sandini, G., & Takanishi, A. (2011). Learning task space control through goal directed exploration. In 2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011 (pp. 702-708). [6181368] https://doi.org/10.1109/ROBIO.2011.6181368

    Learning task space control through goal directed exploration. / Jamone, Lorenzo; Natale, Lorenzo; Hashimoto, Kenji; Sandini, Giulio; Takanishi, Atsuo.

    2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011. 2011. p. 702-708 6181368.

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

    Jamone, L, Natale, L, Hashimoto, K, Sandini, G & Takanishi, A 2011, Learning task space control through goal directed exploration. in 2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011., 6181368, pp. 702-708, 2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011, Phuket, 11/12/7. https://doi.org/10.1109/ROBIO.2011.6181368
    Jamone L, Natale L, Hashimoto K, Sandini G, Takanishi A. Learning task space control through goal directed exploration. In 2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011. 2011. p. 702-708. 6181368 https://doi.org/10.1109/ROBIO.2011.6181368
    Jamone, Lorenzo ; Natale, Lorenzo ; Hashimoto, Kenji ; Sandini, Giulio ; Takanishi, Atsuo. / Learning task space control through goal directed exploration. 2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011. 2011. pp. 702-708
    @inproceedings{64946b84d8264b719e000067c45688b2,
    title = "Learning task space control through goal directed exploration",
    abstract = "We present an autonomous goal-directed strategy to learn how to control a redundant robot in the task space. We discuss the advantages of exploring the state space through goal-directed actions defined in the task space (i.e. learning by trying to do) instead of performing motor babbling in the joints space, and we stress the importance of learning to be performed online, without any separation between training and execution. Our solution relies on learning the forward model and then inverting it for the control; different approaches to learn the forward model are described and compared. Experimental results on a simulated humanoid robot are provided to support our claims. The robot learns autonomously how to perform reaching actions directed toward 3D targets in task space by using arm and waist motion, not relying on any prior knowledge or initial motor babbling. To test the ability of the system to adapt to sudden changes both in the robot structure and in the perceived environment we artificially introduce two different kinds of kinematic perturbations: a modification of the length of one link and a rotation of the task space reference frame. Results demonstrate that the online update of the model allows the robot to cope with such situations.",
    author = "Lorenzo Jamone and Lorenzo Natale and Kenji Hashimoto and Giulio Sandini and Atsuo Takanishi",
    year = "2011",
    doi = "10.1109/ROBIO.2011.6181368",
    language = "English",
    isbn = "9781457721373",
    pages = "702--708",
    booktitle = "2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011",

    }

    TY - GEN

    T1 - Learning task space control through goal directed exploration

    AU - Jamone, Lorenzo

    AU - Natale, Lorenzo

    AU - Hashimoto, Kenji

    AU - Sandini, Giulio

    AU - Takanishi, Atsuo

    PY - 2011

    Y1 - 2011

    N2 - We present an autonomous goal-directed strategy to learn how to control a redundant robot in the task space. We discuss the advantages of exploring the state space through goal-directed actions defined in the task space (i.e. learning by trying to do) instead of performing motor babbling in the joints space, and we stress the importance of learning to be performed online, without any separation between training and execution. Our solution relies on learning the forward model and then inverting it for the control; different approaches to learn the forward model are described and compared. Experimental results on a simulated humanoid robot are provided to support our claims. The robot learns autonomously how to perform reaching actions directed toward 3D targets in task space by using arm and waist motion, not relying on any prior knowledge or initial motor babbling. To test the ability of the system to adapt to sudden changes both in the robot structure and in the perceived environment we artificially introduce two different kinds of kinematic perturbations: a modification of the length of one link and a rotation of the task space reference frame. Results demonstrate that the online update of the model allows the robot to cope with such situations.

    AB - We present an autonomous goal-directed strategy to learn how to control a redundant robot in the task space. We discuss the advantages of exploring the state space through goal-directed actions defined in the task space (i.e. learning by trying to do) instead of performing motor babbling in the joints space, and we stress the importance of learning to be performed online, without any separation between training and execution. Our solution relies on learning the forward model and then inverting it for the control; different approaches to learn the forward model are described and compared. Experimental results on a simulated humanoid robot are provided to support our claims. The robot learns autonomously how to perform reaching actions directed toward 3D targets in task space by using arm and waist motion, not relying on any prior knowledge or initial motor babbling. To test the ability of the system to adapt to sudden changes both in the robot structure and in the perceived environment we artificially introduce two different kinds of kinematic perturbations: a modification of the length of one link and a rotation of the task space reference frame. Results demonstrate that the online update of the model allows the robot to cope with such situations.

    UR - http://www.scopus.com/inward/record.url?scp=84860720009&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84860720009&partnerID=8YFLogxK

    U2 - 10.1109/ROBIO.2011.6181368

    DO - 10.1109/ROBIO.2011.6181368

    M3 - Conference contribution

    SN - 9781457721373

    SP - 702

    EP - 708

    BT - 2011 IEEE International Conference on Robotics and Biomimetics, ROBIO 2011

    ER -