Stefan Schaal
Georgia Institute of Technology, University of Southern California, Southern California University for Professional Studies, Research Organization of Information and Systems, Computational Physics (United States), Massachusetts Institute of Technology, Advanced Telecommunications Research Institute International, Japan Science and Technology Agency, Max Planck Institute for Intelligent Systems, Max Planck Society, Google (United States), Intrinsic LifeSciences (United States), École Polytechnique Fédérale de Lausanne, Technische Universität Darmstadt, Tohoku University, LAC+USC Medical Center
Papers
154
Total Citations
16,746
H-Index
58
About
Stefan Schaal is a pioneering researcher at the intersection of robotics, machine learning, and computational neuroscience, whose work has fundamentally shaped how robots learn and execute motor skills. Best known for his groundbreaking contributions to **imitation learning**, **reinforcement learning**, and **movement primitives**, Schaal has developed frameworks that bridge human motor behavior and robotic control in transformative ways. His early work on locally weighted learning (1997, 1,683 citations) established powerful non-parametric methods for function approximation that became foundational across machine learning. He subsequently advanced robot programming by demonstration (984 citations), enabling robots to acquire complex behaviors by observing human movements rather than requiring explicit programming. His development of Dynamic Movement Primitives — explored across several highly cited works including "Learning Attractor Landscapes" (557 citations) and "Learning Movement Primitives" (346 citations) — provided elegant mathematical representations for encoding and generalizing motor skills. Schaal's contributions to policy gradient methods and reinforcement learning for robotics, including the Natural Actor-Critic framework (751 citations) and path integral control approaches (449 citations), have given the field scalable, practical algorithms for continuous control. More recently, his work on Time-Contrastive Networks (555 citations) demonstrates his reach into self-supervised learning from video. With tens of thousands of cumulative citations, Schaal's legacy spans both theoretical rigor and real-world robotic application.
Research Focus
Key Achievements
Top Papers
- 1Locally Weighted Learning1,683 citations · 1997
- 2Robot Programming by Demonstration984 citations · 2008
- 3Reinforcement learning of motor skills with policy gradients863 citations · 2008
- 4Natural Actor-Critic751 citations · 2008
- 5Learning and generalization of motor skills by learning from demonstration710 citations · 2009
- 6Learning Attractor Landscapes for Learning Motor Primitives557 citations · 2002
- 7Time-Contrastive Networks: Self-Supervised Learning from Video555 citations · 2018
- 8Policy Gradient Methods for Robotics519 citations · 2006
- 9A Generalized Path Integral Control Approach to Reinforcement Learning449 citations · 2010
- 10Learning Movement Primitives346 citations · 2005