Sebastian Blaes

Max Planck Institute for Intelligent Systems, Max Planck Society

Papers

Total Citations

H-Index

About

Sebastian Blaes is a robotics researcher advancing the frontier of skill acquisition and autonomous learning. His work centers on enabling robots to learn diverse, agile behaviors without the need for meticulously labeled datasets or hand-crafted reward functions. A key contribution is his development of self-supervised adversarial imitation learning, which allows robots to extract and control a wide range of skills from unlabeled, mixed motion data, as detailed in his 2023 paper (19 citations). Blaes has also pioneered methods for learning agile skills from rough, partial demonstrations, significantly reducing the burden of providing perfect expert examples. Recognizing the critical gap between simulation and reality, he co-authored a benchmark for offline reinforcement learning on real-robot hardware (11 citations), providing a vital framework for the community. His earlier work on intrinsically motivated task-planning agents explores how robots can autonomously decide what to learn and how to allocate attention, a foundational concept for open-ended learning. With a growing citation impact, Blaes’s research is shaping a future where robots can learn complex, real-world skills with minimal human supervision.