Petar Kormushev

About

Petar Kormushev is a prominent robotics researcher whose work spans reinforcement learning, imitation learning, and humanoid robot control. He has made foundational contributions to how robots acquire and refine motor skills, particularly through his pioneering methods in kinesthetic teaching — enabling robots to learn complex force interactions directly from human demonstrations. His 2011 paper on imitation learning of positional and force skills (277 citations) and his EM-based reinforcement learning approach for motor skill coordination (265 citations) represent landmark advances in robot learning from human guidance. Kormushev's widely cited 2013 survey on reinforcement learning in robotics (253 citations) has become an essential reference for researchers navigating real-world applications and challenges in the field. His work extends across bipedal locomotion — including energy-efficient walking through reinforcement learning and online gait regeneration — as well as tactile sensing for object pose estimation and symbolic action representation learning. Perhaps most strikingly, his research demonstrated that a humanoid iCub robot could autonomously learn archery, illustrating his talent for translating sophisticated learning algorithms into compelling real-world robotic achievements. Across his body of work, Kormushev has consistently bridged theoretical machine learning methods and practical robotic implementation.

Research Focus

Key Achievements

Top Papers

1. 1
   Imitation Learning of Positional and Force Skills Demonstrated via Kinesthetic Teaching and Haptic Input

   277 citations · 2011
2. 2
   Robot motor skill coordination with EM-based Reinforcement Learning

   265 citations · 2010
3. 3
   Reinforcement Learning in Robotics: Applications and Real-World Challenges

   253 citations · 2013
4. 4
   Learning the skill of archery by a humanoid robot iCub

   72 citations · 2010
5. 5
   Online regeneration of bipedal walking gait pattern optimizing footstep placement and timing

   67 citations · 2015
6. 6
   Upper-body kinesthetic teaching of a free-standing humanoid robot

   66 citations · 2011
7. 7
   Learning symbolic representations of actions from human demonstrations

   65 citations · 2015
8. 8
   Bipedal walking energy minimization by reinforcement learning with evolving policy parameterization

   60 citations · 2011
9. 9
   Compliant skills acquisition and multi-optima policy search with EM-based reinforcement learning

   55 citations · 2012
10. 10
    Global estimation of an object’s pose using tactile sensing

    45 citations · 2015

Key Collaborators

Contact & Links