Maximize-perturb-minimize: A fast and effective heuristic to obtain sets of locally optimal robot postures

Martim Brandao, Kenji Hashimoto, Atsuo Takanishi

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

    Abstract

    Complex robots such as legged and humanoid robots are often characterized by non-convex optimization landscapes with multiple local minima. Obtaining sets of these local minima has interesting applications in global optimization, as well as in smart teleoperation interfaces with automatic posture suggestions. In this paper we propose a new heuristic method to obtain sets of local minima, which is to run multiple minimization problems initialized around a local maximum. The method is simple, fast, and produces diverse postures from a single nominal posture. Results on the robot WAREC-1 using a sum-of-squared-Torques cost function show that our method quickly obtains lower-cost postures than typical random restart strategies. We further show that obtained postures are more diverse than when sampling around nominal postures, and that they are more likely to be feasible when compared to a uniform-sampling strategy. We also show that lack of completeness leads to the method being most useful when computation has to be fast, but not on very large computation time budgets.

    Original languageEnglish
    Title of host publication2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017
    PublisherInstitute of Electrical and Electronics Engineers Inc.
    Pages2624-2629
    Number of pages6
    Volume2018-January
    ISBN (Electronic)9781538637418
    DOIs
    Publication statusPublished - 2018 Mar 23
    Event2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017 - Macau, China
    Duration: 2017 Dec 52017 Dec 8

    Other

    Other2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017
    CountryChina
    CityMacau
    Period17/12/517/12/8

    Fingerprint

    Local Minima
    Robot
    Maximise
    Heuristics
    Robots
    Minimise
    Categorical or nominal
    Sampling
    Legged Robots
    Teleoperation
    Sampling Strategy
    Humanoid Robot
    Nonconvex Optimization
    Heuristic methods
    Restart
    Heuristic Method
    Global optimization
    Remote control
    Cost functions
    Global Optimization

    ASJC Scopus subject areas

    • Artificial Intelligence
    • Mechanical Engineering
    • Control and Optimization
    • Modelling and Simulation

    Cite this

    Brandao, M., Hashimoto, K., & Takanishi, A. (2018). Maximize-perturb-minimize: A fast and effective heuristic to obtain sets of locally optimal robot postures. In 2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017 (Vol. 2018-January, pp. 2624-2629). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ROBIO.2017.8324815

    Maximize-perturb-minimize : A fast and effective heuristic to obtain sets of locally optimal robot postures. / Brandao, Martim; Hashimoto, Kenji; Takanishi, Atsuo.

    2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017. Vol. 2018-January Institute of Electrical and Electronics Engineers Inc., 2018. p. 2624-2629.

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

    Brandao, M, Hashimoto, K & Takanishi, A 2018, Maximize-perturb-minimize: A fast and effective heuristic to obtain sets of locally optimal robot postures. in 2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017. vol. 2018-January, Institute of Electrical and Electronics Engineers Inc., pp. 2624-2629, 2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017, Macau, China, 17/12/5. https://doi.org/10.1109/ROBIO.2017.8324815
    Brandao M, Hashimoto K, Takanishi A. Maximize-perturb-minimize: A fast and effective heuristic to obtain sets of locally optimal robot postures. In 2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017. Vol. 2018-January. Institute of Electrical and Electronics Engineers Inc. 2018. p. 2624-2629 https://doi.org/10.1109/ROBIO.2017.8324815
    Brandao, Martim ; Hashimoto, Kenji ; Takanishi, Atsuo. / Maximize-perturb-minimize : A fast and effective heuristic to obtain sets of locally optimal robot postures. 2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017. Vol. 2018-January Institute of Electrical and Electronics Engineers Inc., 2018. pp. 2624-2629
    @inproceedings{f6ae5c359f1249ada23e27599e89111f,
    title = "Maximize-perturb-minimize: A fast and effective heuristic to obtain sets of locally optimal robot postures",
    abstract = "Complex robots such as legged and humanoid robots are often characterized by non-convex optimization landscapes with multiple local minima. Obtaining sets of these local minima has interesting applications in global optimization, as well as in smart teleoperation interfaces with automatic posture suggestions. In this paper we propose a new heuristic method to obtain sets of local minima, which is to run multiple minimization problems initialized around a local maximum. The method is simple, fast, and produces diverse postures from a single nominal posture. Results on the robot WAREC-1 using a sum-of-squared-Torques cost function show that our method quickly obtains lower-cost postures than typical random restart strategies. We further show that obtained postures are more diverse than when sampling around nominal postures, and that they are more likely to be feasible when compared to a uniform-sampling strategy. We also show that lack of completeness leads to the method being most useful when computation has to be fast, but not on very large computation time budgets.",
    author = "Martim Brandao and Kenji Hashimoto and Atsuo Takanishi",
    year = "2018",
    month = "3",
    day = "23",
    doi = "10.1109/ROBIO.2017.8324815",
    language = "English",
    volume = "2018-January",
    pages = "2624--2629",
    booktitle = "2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017",
    publisher = "Institute of Electrical and Electronics Engineers Inc.",

    }

    TY - GEN

    T1 - Maximize-perturb-minimize

    T2 - A fast and effective heuristic to obtain sets of locally optimal robot postures

    AU - Brandao, Martim

    AU - Hashimoto, Kenji

    AU - Takanishi, Atsuo

    PY - 2018/3/23

    Y1 - 2018/3/23

    N2 - Complex robots such as legged and humanoid robots are often characterized by non-convex optimization landscapes with multiple local minima. Obtaining sets of these local minima has interesting applications in global optimization, as well as in smart teleoperation interfaces with automatic posture suggestions. In this paper we propose a new heuristic method to obtain sets of local minima, which is to run multiple minimization problems initialized around a local maximum. The method is simple, fast, and produces diverse postures from a single nominal posture. Results on the robot WAREC-1 using a sum-of-squared-Torques cost function show that our method quickly obtains lower-cost postures than typical random restart strategies. We further show that obtained postures are more diverse than when sampling around nominal postures, and that they are more likely to be feasible when compared to a uniform-sampling strategy. We also show that lack of completeness leads to the method being most useful when computation has to be fast, but not on very large computation time budgets.

    AB - Complex robots such as legged and humanoid robots are often characterized by non-convex optimization landscapes with multiple local minima. Obtaining sets of these local minima has interesting applications in global optimization, as well as in smart teleoperation interfaces with automatic posture suggestions. In this paper we propose a new heuristic method to obtain sets of local minima, which is to run multiple minimization problems initialized around a local maximum. The method is simple, fast, and produces diverse postures from a single nominal posture. Results on the robot WAREC-1 using a sum-of-squared-Torques cost function show that our method quickly obtains lower-cost postures than typical random restart strategies. We further show that obtained postures are more diverse than when sampling around nominal postures, and that they are more likely to be feasible when compared to a uniform-sampling strategy. We also show that lack of completeness leads to the method being most useful when computation has to be fast, but not on very large computation time budgets.

    UR - http://www.scopus.com/inward/record.url?scp=85049913904&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85049913904&partnerID=8YFLogxK

    U2 - 10.1109/ROBIO.2017.8324815

    DO - 10.1109/ROBIO.2017.8324815

    M3 - Conference contribution

    AN - SCOPUS:85049913904

    VL - 2018-January

    SP - 2624

    EP - 2629

    BT - 2017 IEEE International Conference on Robotics and Biomimetics, ROBIO 2017

    PB - Institute of Electrical and Electronics Engineers Inc.

    ER -