Pavel Kolev
Papers
3
Total Citations
33
H-Index
3
About
Pavel Kolev is a robotics researcher whose work sits at the intersection of imitation learning, reinforcement learning, and autonomous skill acquisition. His research addresses one of the field's most persistent challenges: enabling robots to learn and control diverse, meaningful behaviors from complex, real-world data. Kolev's most recognized contribution, "Versatile Skill Control via Self-supervised Adversarial Imitation of Unlabeled Mixed Motions" (2023, 19 citations), introduced a framework allowing robots to learn distinct skills from unlabeled, heterogeneous motion datasets — removing the costly requirement for manually annotated data. This work represents a significant step toward scalable robot learning in unstructured environments. His concurrent work on benchmarking offline reinforcement learning on real robot hardware (2023, 11 citations) advances the practical side of the field, rigorously evaluating how data-driven policies transfer to physical dexterous manipulation systems — a notoriously difficult open problem. His more recent work on diverse skill learning under multi-constraint optimality (2024) further refines how robots can balance competing task requirements without sacrificing behavioral diversity. Collectively, Kolev's research pushes the boundaries of sample-efficient, scalable robot learning, making him a notable emerging voice in modern robotics and embodied AI.
Research Focus
Key Achievements
Top Papers
- 1
- 2Benchmarking Offline Reinforcement Learning on Real-Robot Hardware11 citations · 2023
- 3