Michiel van de Panne
Papers
18
Total Citations
1,451
H-Index
11
About
Michiel van de Panne is a leading figure in computer animation and robotics, renowned for pioneering the use of deep reinforcement learning (RL) to create highly dynamic and robust legged locomotion. His work bridges the gap between simulated characters and real-world robots, with a focus on bipedal and quadrupedal movement. His landmark 2018 paper, "DeepMimic" (802 citations), revolutionized character animation by showing how RL could imitate motion capture data in physics simulation, enabling realistic responses to perturbations. He extended this to hardware with "Feedback Control For Cassie With Deep Reinforcement Learning" (188 citations), demonstrating that learned policies could control a real bipedal robot. Van de Panne’s contributions include curriculum-driven learning for complex tasks like stepping-stone locomotion ("ALLSTEPS," 99 citations) and systematic studies on sim-to-real transfer ("Dynamics Randomization Revisited," 75 citations). His work on the Cassie robot (63 citations) and generalizable quadrupedal locomotion ("GLiDE," 43 citations) has set benchmarks in the field. With over 1,300 citations across his top papers, van de Panne has shaped modern legged robotics and animation, and his recent state-of-the-art review (2025) underscores his ongoing influence.
Research Focus
Key Achievements
Top Papers
- 1DeepMimic802 citations · 2018
- 2Feedback Control For Cassie With Deep Reinforcement Learning188 citations · 2018
- 3ALLSTEPS: Curriculum‐driven Learning of Stepping Stone Skills99 citations · 2020
- 4Dynamics Randomization Revisited: A Case Study for Quadrupedal Locomotion75 citations · 2021
- 5
- 6
- 7
- 8Learning-based legged locomotion: State of the art and future perspectives34 citations · 2025
- 9
- 10Feedback Control For Cassie With Deep Reinforcement Learning21 citations · 2018