Takamitsu MATSUBARA


Ph. D. Candidate
Computational Neuroscience Lab(at ATR)
Graduate School of Information Science
Nara Institute of Science and Technology

Research Fellow (DC1, April 2005 - March 2008)
The Japan Society for the Promotion Science (JSPS)

---

Research Interests

• Reinforcement learning
• Biped Locomotion
• Bayesian Inference
• Statistical Learning
• Control Theory
• Whole-body control of humanoid robot

---

Awards

• IEEE Kansai Section Student Award, February (2006)

---

Publications

• Internatinal Journals
  • Learning CPG-based Biped Locomotion with a Policy Gradient Method
    Takamitsu Matsubara, Jun Morimoto, Jun Nakanishi, Masa-akiSato, and Kenji Doya
    Robotics and Autonomus Systems, vol. 54, issue 11, pp. 911-920 (2006)

• Internatinal Conferences
  
  • Learning to acquire whole-body humanoid CoM movements to achieve dynamic tasks
    Takamitsu Matsubara, Jun Morimoto, Jun Nakanishi, Sang-Ho Hyon, Joshua G.Hale and Gordon Cheng
    IEEE International Conference on Robotics and Automation, Italy (Accepted)
  • Learning to acquire whole-body humanoid CoM movements to achieve dynamic tasks with a policy gradient method
    Takamitsu Matsubara, Jun Morimoto, Jun Nakanishi, Sang-Ho Hyon, Joshua G.Hale and Gordon Cheng
    Advances in Neural Information Processing Systems (NIPS '06 Workshop), Canada.
  • Learning CPG-based biped locomotion with a Policy Gradient Method
    Takamitsu Matsubara, Jun Morimoto, Jun Matsubara, Masa-aki Sato, Kenji Doya
    IEEE-RAS International Conference on Humanoids Robots Humanoids2005, pp. 208-213, Tsukuba, Japan (2005)
  • Learning CPG Sensory Feedback with Policy Gradient for Biped Locomotion for a Full-body Humanoid
    Gen Endo, Jun Morimoto, Takamitsu Matsubara, Jun Nakanishi, and Gordon Cheng
    The Twentieth National Conference on Artificial Intelligence, pp. 1267-1273 (AAAI-05), Pittsburgh, PA, USA, July 9--13 (2005)
  • Learning Sensory Feedback to CPG with Policy Gradient for Biped Locomotion
    Takamitsu Matsubara, Jun Morimoto, Jun Nakanishi, Masa-aki Sato, and Kenji Doya
    IEEE International Conference on Robotics and Automation, pp. 4175-4180, Barcelona, Spain, April 18-22 (2005)

• Domestic Journals
  • 方策こう配法を用いた動的行動則の獲得：2足歩行への適用
    松原崇充, 森本 淳, 中西　淳，佐藤雅昭，銅谷賢治
    電子情報通信学会論文誌 vol. J88-D2, no. 1, pp. 53-65 (2005)

• Domestic Conferences
  • 変分ベイズ法による自然方策勾配の推定法
    松原崇充, 森本 淳, 中西　淳，佐藤雅昭，銅谷賢治
    電子情報通信学会技術研究報告，NC2005-52, pp. 37-42 (2005-10)
  • Learning sensory feedback to CPG with a policy gradient method and its application ot a real biped robot
    Takamitsu Matsubara, Jun Morimoto, Jun Nakanishi, Masa-aki Sato, Kenji Doya
    脳と心のメカニズム 第6回 夏のワークショップ
  • 2足歩行運動のための動的行動則の獲得
    松原崇充，森本　淳，中西　淳，佐藤雅昭，銅谷賢治
    日本機械学会ロボティクス・メカトロニクス講演会, 1A1-L1-51 (CD-ROM) (2004)
  • 方策勾配法を用いた動的行動則の獲得：2足歩行への適用
    松原崇充, 森本 淳, 中西　淳，佐藤雅昭，銅谷賢治
    電子情報通信学会技術研究報告，NC2003-128, pp. 53-58 (2004-1)

---

Movies

• Learning biped walking
  
  • DB-chan's CPG-based biped walking [before learning], [after learning], [walking on a seesaw-like metal sheet]
  • Hoap2's simple CPG-walking [CPG-walking],
  
  
• Learning whole-body movement
  
  • Hoap2's dynamic whole-body punch [before learning], [after learning],