Manabu Nishiura

This page introduces my projects.

Home About Projects cv

Researches

Human Coexitance Robots and its Learning-based Control

In order to realize a robot that can perform housework in the same environment and at the same or faster speed as a human, we have designed a robot with low inertia and high backdrivability and shown that learning-based control is feasible.
  1. Development of Low-inertia Backdribable Arm Focusing on Learning-based Control
    Manabu Nishiura, Akira Hanano, Kazutoshi Nishii, Yoshihiro Okumatsu
    to be presented at The 2022 IEEE-RAS International Conference on Intelligent Robots and Systems (IROS2022)

Learning-based Control of Contact-rich Motion in Musculoskeletal Humanoid

I studied a system that enables robots to learn environmental contact behavior on their own, referring to the hierarchical structure of the human brain. By combining reinforcement learning and self-supervised learning, a musculoskeletal humanoid succeeded in performing a sitting motion on a car seat.
  1. ロボット身体のための適応的な環境接触行動における継続的自己学習のための階層型学習システムの研究
    西浦 学，岡田慧，稲葉雅幸
    東京大学大学院学際情報学府学際情報学専攻先端表現情報学コース
    Slides
  2. 接触状態を含む身体モデルと強化学習を用いた筋骨格ヒューマノイドによる環境接触行動
    西浦 学，河原塚健人，利光泰徳，浅野悠紀，岡田慧，川崎宏治，稲葉雅幸
    東京大学大学院学際情報学府学際情報学専攻先端表現情報学コース
    Paper (in Japanese)
  3. Development of Musculoskeletal Legs with Planar Interskeletal Structures to Realize Human Comparable Moving Function.
    Onitsuka, Moritaka; Nishiura, Manabu; Kawaharazuka, Kento; Tsuzuki, Kei; Toshimitsu, Yasunori; Omura, Yusuke; Asano, Yuki; Okada, Kei; Kawasaki, Koji; Inaba, Masayuki
    The 2020 IEEE-RAS International Conference on Humanoid Robots (HUMANOIDS2020)
    Paper (in Japanese)
  4. Automatic Grouping of Redundant Sensors and Actuators Using Functional and Spatial Connections: Application to Muscle Grouping for Musculoskeletal Humanoids.
    Kento Kawaharazuka, Manabu Nishiura, Yuya Koga, Yusuke Omura, Yasunori Toshimitsu, Yuki Asano, Kei Okada, Koji Kawasaki, Masayuki Inaba
    IEEE Robotics and Automation Letters
    Paper Link
  5. Stability Recognition with Active Vibration for Bracing Behaviors and Motion Extensions Using Environment in Musculoskeletal Humanoids.
    Kento Kawaharazuka, Manabu Nishiura, Shinsuke Nakashima, Yasunori Toshimitsu, Yusuke Omura, Yuya Koga, Yuki Asano, Kei Okada, Koji Kawasaki, Masayuki Inaba
    Proceedings of the 2021 IEEE International Conference on Soft Robotics
    Paper Link
  6. Biomimetic Operational Space Control for Musculoskeletal Humanoid Optimizing Across Muscle Activation and Joint Nullspace.
    Y. Toshimitsu, K. Kawaharazuka, M. Nishiura, Y. Koga, Y. Omura, Y. Asano, K. Okada, K. Kawasaki, and M. Inaba
    2021 International Conference on Robotics and Automation (ICRA 2021)
    Paper Link

Conputational Neuroscience

Learned nonlinear dynamics and modeled the behavior of the pupulation dynamics of neurons. Proposed the model which can replication the temperal and spacial patterns of the population of neurons.
  1. 微小電極アレイ基板上の神経細胞集団の同期バーストを評価するマルチスケールモデル
    西浦 学，赤尾旭彦，榛葉健太，小谷潔，神保泰彦
    電気学会C部門研究会，医用生体工学研究会,2019,東京大学先端科学技術センター
    Paper (in Japanese) Google Slides

Generation of Motion using Deep Reinforcement Learning

Generate policy which can handle hard tasks using deep reinforcement learning.
  1. Replication of "Emergence of Locomotion Behaviors in Rich Environement"
    research blog

Acquisition of Motor Intelligence in Musculoskeletal Humanoids

Human body has many merits. For example, the elasticity of tendon can absorb impact forces. In addition, redundant actuators enable variable stiffness control of work space.
  1. 筋骨格ヒューマノイドにおける環境物体に応じた適応的剛性レンジ選択とその可変剛性制御戦略の獲得
    西浦 学，河原塚 健人， 鬼塚 盛宇， 浅野 悠紀， 岡田 慧，稲葉 雅幸
    日本ロボット学会，2019，早稲田大学
    Google Slides

Autonomous Driving Experiment with the Humanoid Robot in Toyota City: Collaborative Research with Toyota Inc

Taking advantage of the human mimetic musculoskeletal humanoids’ mechanism which can fit to the environment, we are doing research on functions of both the hardware and the information processing of a human being which enable a human being to do a series of action related to driving. I am in charge of development of new legs and its controller which can switch pedal with single leg like human-being.
  1. press1, press2, movie

  1. Toward Autonomous Driving by Musculoskeletal Humanoids: A Study of Developed Hardware and Learning-Based Software
    Kento Kawaharazuka, Kei Tsuzuki, Yuya Koka, Yusuke Omura, Tasuku Makebe, Koki Shinjo, Moritaka Onitsuka, Yuya Nagamatsu, Yuki Asano, Kei Okada, Koji Kawasaki, Masayuki Inaba
    in IEEE Robotics & Automation Magazine, vol.27, no.3, pp. 84-96, Sept. 2020
    RA Magazine

  1. 面状骨格間構造を利用し広い可動域においてモーメントアームを確保し高出力での環境接触動作が可能な筋骨格脚の開発
    鬼塚 盛宇，○西浦 学，河原塚 健人，都築 敬，利光 泰徳，大村 柚介，浅野 悠紀，岡田 慧，川崎 宏治，稲葉 雅幸
    日本ロボット学会，2020，オンライン

Other Projects

Flying Robot Project

Designed and succeeded in flying tail-less flying robot (~ 1m, ~ 200g).
  1. Fisrt tail-less flying robot which could fly
    videos and pictures

Data-driven Science

The growing model of leafvein was chosen based on the previous research. First state (distribution of hormone substances and raw materials of leafvein)is setted under the assumption. Final state is the real world data (real leafvein data). Parameters is chosen so that the first state can evolve to the final state.
  1. gif, code and presentation