Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior

Shingo Murata, Wataru Masuda, Saki Tomioka, Tetsuya Ogata, Shigeki Sugano

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

    2 Citations (Scopus)

    Abstract

    In this paper, we propose a method to dynamically modulate the input state of recurrent neural networks (RNNs) so as to realize flexible and robust robot behavior. We employ the so-called stochastic continuous-time RNN (S-CTRNN), which can learn to predict the mean and variance (or uncertainty) of subsequent sensorimotor information. Our proposed method uses this estimated uncertainty to determine a mixture ratio for combining actual and predicted sensory states of network input. The method is evaluated by conducting a robot learning experiment in which a robot is required to perform a sensory-dependent task and a sensory-independent task. The sensory-dependent task requires the robot to incorporate meaningful sensory information, and the sensory-independent task requires the robot to ignore irrelevant sensory information. Experimental results demonstrate that a robot controlled by our proposed method exhibits flexible and robust behavior, which results from dynamic modulation of the network input on the basis of the estimated uncertainty of actual sensory states.

    Original languageEnglish
    Title of host publicationArtificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings
    PublisherSpringer-Verlag
    Pages11-18
    Number of pages8
    ISBN (Print)9783319685991
    DOIs
    Publication statusPublished - 2017 Jan 1
    Event26th International Conference on Artificial Neural Networks, ICANN 2017 - Alghero, Italy
    Duration: 2017 Sep 112017 Sep 14

    Publication series

    NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
    Volume10613 LNCS
    ISSN (Print)0302-9743
    ISSN (Electronic)1611-3349

    Other

    Other26th International Conference on Artificial Neural Networks, ICANN 2017
    CountryItaly
    CityAlghero
    Period17/9/1117/9/14

    Fingerprint

    Uncertainty Estimation
    Robot
    Robots
    Recurrent neural networks
    Recurrent Neural Networks
    Uncertainty
    Robot learning
    Dependent
    Modulation
    Continuous Time
    Predict
    Experimental Results
    Demonstrate
    Experiments
    Experiment

    Keywords

    • Neurorobotics
    • Recurrent neural networks
    • Robot
    • Uncertainty

    ASJC Scopus subject areas

    • Theoretical Computer Science
    • Computer Science(all)

    Cite this

    Murata, S., Masuda, W., Tomioka, S., Ogata, T., & Sugano, S. (2017). Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior. In Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings (pp. 11-18). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10613 LNCS). Springer-Verlag. https://doi.org/10.1007/978-3-319-68600-4_2

    Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior. / Murata, Shingo; Masuda, Wataru; Tomioka, Saki; Ogata, Tetsuya; Sugano, Shigeki.

    Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings. Springer-Verlag, 2017. p. 11-18 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10613 LNCS).

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

    Murata, S, Masuda, W, Tomioka, S, Ogata, T & Sugano, S 2017, Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior. in Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10613 LNCS, Springer-Verlag, pp. 11-18, 26th International Conference on Artificial Neural Networks, ICANN 2017, Alghero, Italy, 17/9/11. https://doi.org/10.1007/978-3-319-68600-4_2
    Murata S, Masuda W, Tomioka S, Ogata T, Sugano S. Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior. In Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings. Springer-Verlag. 2017. p. 11-18. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-68600-4_2
    Murata, Shingo ; Masuda, Wataru ; Tomioka, Saki ; Ogata, Tetsuya ; Sugano, Shigeki. / Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior. Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings. Springer-Verlag, 2017. pp. 11-18 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
    @inproceedings{4ed1a0dd9bf04039bb07164245534fc3,
    title = "Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior",
    abstract = "In this paper, we propose a method to dynamically modulate the input state of recurrent neural networks (RNNs) so as to realize flexible and robust robot behavior. We employ the so-called stochastic continuous-time RNN (S-CTRNN), which can learn to predict the mean and variance (or uncertainty) of subsequent sensorimotor information. Our proposed method uses this estimated uncertainty to determine a mixture ratio for combining actual and predicted sensory states of network input. The method is evaluated by conducting a robot learning experiment in which a robot is required to perform a sensory-dependent task and a sensory-independent task. The sensory-dependent task requires the robot to incorporate meaningful sensory information, and the sensory-independent task requires the robot to ignore irrelevant sensory information. Experimental results demonstrate that a robot controlled by our proposed method exhibits flexible and robust behavior, which results from dynamic modulation of the network input on the basis of the estimated uncertainty of actual sensory states.",
    keywords = "Neurorobotics, Recurrent neural networks, Robot, Uncertainty",
    author = "Shingo Murata and Wataru Masuda and Saki Tomioka and Tetsuya Ogata and Shigeki Sugano",
    year = "2017",
    month = "1",
    day = "1",
    doi = "10.1007/978-3-319-68600-4_2",
    language = "English",
    isbn = "9783319685991",
    series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
    publisher = "Springer-Verlag",
    pages = "11--18",
    booktitle = "Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings",

    }

    TY - GEN

    T1 - Mixing actual and predicted sensory states based on uncertainty estimation for flexible and robust robot behavior

    AU - Murata, Shingo

    AU - Masuda, Wataru

    AU - Tomioka, Saki

    AU - Ogata, Tetsuya

    AU - Sugano, Shigeki

    PY - 2017/1/1

    Y1 - 2017/1/1

    N2 - In this paper, we propose a method to dynamically modulate the input state of recurrent neural networks (RNNs) so as to realize flexible and robust robot behavior. We employ the so-called stochastic continuous-time RNN (S-CTRNN), which can learn to predict the mean and variance (or uncertainty) of subsequent sensorimotor information. Our proposed method uses this estimated uncertainty to determine a mixture ratio for combining actual and predicted sensory states of network input. The method is evaluated by conducting a robot learning experiment in which a robot is required to perform a sensory-dependent task and a sensory-independent task. The sensory-dependent task requires the robot to incorporate meaningful sensory information, and the sensory-independent task requires the robot to ignore irrelevant sensory information. Experimental results demonstrate that a robot controlled by our proposed method exhibits flexible and robust behavior, which results from dynamic modulation of the network input on the basis of the estimated uncertainty of actual sensory states.

    AB - In this paper, we propose a method to dynamically modulate the input state of recurrent neural networks (RNNs) so as to realize flexible and robust robot behavior. We employ the so-called stochastic continuous-time RNN (S-CTRNN), which can learn to predict the mean and variance (or uncertainty) of subsequent sensorimotor information. Our proposed method uses this estimated uncertainty to determine a mixture ratio for combining actual and predicted sensory states of network input. The method is evaluated by conducting a robot learning experiment in which a robot is required to perform a sensory-dependent task and a sensory-independent task. The sensory-dependent task requires the robot to incorporate meaningful sensory information, and the sensory-independent task requires the robot to ignore irrelevant sensory information. Experimental results demonstrate that a robot controlled by our proposed method exhibits flexible and robust behavior, which results from dynamic modulation of the network input on the basis of the estimated uncertainty of actual sensory states.

    KW - Neurorobotics

    KW - Recurrent neural networks

    KW - Robot

    KW - Uncertainty

    UR - http://www.scopus.com/inward/record.url?scp=85034265361&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85034265361&partnerID=8YFLogxK

    U2 - 10.1007/978-3-319-68600-4_2

    DO - 10.1007/978-3-319-68600-4_2

    M3 - Conference contribution

    SN - 9783319685991

    T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

    SP - 11

    EP - 18

    BT - Artificial Neural Networks and Machine Learning – ICANN 2017 - 26th International Conference on Artificial Neural Networks, Proceedings

    PB - Springer-Verlag

    ER -