Robot Learning

Research Staff

  • Assoc.Prof. Takamitsu Matsubara

    Assoc.Prof.
    Takamitsu Matsubara

E-mail { takam-m }[at] is.naist.jp

Research Areas

Machine learning algorithms for real world robots

(Deep) reinforcement learning

(Deep) imitation learning

Deep learning for dynamical systems

Active perception

Human-in-the-loop optimization

Real world applications

Smart manufacturing

Human-assistive technology (exoskeleton robots, EMG interface etc.)

Chemical plant modeling and control

Vehicle autopiloting

Research equipment

Nextage robot (Kawada)

Baxter robot (Rethink)

UR5 and UR3 (Universal robots)

OP3 humanoid robot (Robotis)

Various sensors (motion capture systems, EMG sensors, etc.)

Collaborators

University of Technology Sydney (Australia), Radboud Univ. (The Netherland), Karlsruhe Institute of Technology (Germany), Edinburgh Univ. (UK), LAAS-CNRS (France), ATR, AIST, Shinshu Univ., Ritsumeikan Univ. Kansai Univ. (Japan), etc.

Research Statement

Robot learning (machine learning for robots) is an interdisciplinary field of various fields such as machine learning, artificial intelligence, robot engineering, control engineering, signal processing, optimization, and mechatronics. You may be able to find your approach by utilizing your field of expertise, skills, and experience (robot contests, programming contests, work, etc.). Please challenge yourself within robot learning research.

Fig.1 Deep reinforcement learning for cloth manipulation

Fig.1 Deep reinforcement learning for cloth manipulation

Fig.2 Object search with Gaussian processes

Fig.2 Object search with Gaussian processes

Fig.3 Object shape estimation from tactile sensing

Fig.3 Object shape estimation from tactile sensing