Takamitsu Matsubara

About

Takamitsu Matsubara is a prominent robotics and machine learning researcher whose work spans biped locomotion, myoelectric interfaces, robotic manipulation, and assistive exoskeleton systems. He has made foundational contributions to the intersection of reinforcement learning and physical robotics, tackling some of the most challenging problems in embodied AI. Matsubara's early landmark work on CPG-based biped locomotion using policy gradient methods — accumulating over 300 citations across two related papers — demonstrated how humanoid robots could acquire stable walking behaviors through principled learning frameworks. His development of Stylistic and Parametric Dynamic Movement Primitives advanced the field's understanding of how robots can generalize motion from multiple human demonstrations. In the realm of human-machine interaction, his bilinear EMG modeling approach (181 citations) enabled user-independent myoelectric interfaces, a significant step toward practical prosthetic and robotic control. His deep reinforcement learning contributions, particularly smooth policy update methods for cloth manipulation (179 citations) and force control for rigid robots (150 citations), have become widely referenced benchmarks. Matsubara's research on clothing assistance and exoskeleton learning further underscores his commitment to socially impactful robotics, addressing real-world needs in elderly care and rehabilitation. His body of work, totaling over 1,200 citations, reflects a career dedicated to making robots genuinely useful in human environments.

Research Focus

Key Achievements

Top Papers

1. 1
   Learning CPG-based Biped Locomotion with a Policy Gradient Method: Application to a Humanoid Robot

   211 citations · 2008
2. 2
   Bilinear Modeling of EMG Signals to Extract User-Independent Features for Multiuser Myoelectric Interface

   181 citations · 2013
3. 3
   Deep reinforcement learning with smooth policy update: Application to robotic cloth manipulation

   179 citations · 2018
4. 4
   Learning Force Control for Contact-Rich Manipulation Tasks With Rigid Position-Controlled Robots

   150 citations · 2020
5. 5
   Reinforcement learning of clothing assistance with a dual-arm robot

   100 citations · 2011
6. 6
   Learning parametric dynamic movement primitives from multiple demonstrations

   99 citations · 2011
7. 7
   Learning CPG-based biped locomotion with a policy gradient method

   98 citations · 2006
8. 8
   Learning assistive strategies for exoskeleton robots from user-robot physical interaction

   79 citations · 2017
9. 9
   XoR: Hybrid drive exoskeleton robot that can balance

   75 citations · 2011
10. 10
    Learning Stylistic Dynamic Movement Primitives from multiple demonstrations

    53 citations · 2010

Key Collaborators

Contact & Links