Papers
22
Total Citations
975
H-Index
13
About
Glen Berseth is a prominent robotics and machine learning researcher whose work sits at the intersection of deep reinforcement learning, legged locomotion, and large-scale robot learning. He is perhaps best known for his foundational contributions to bipedal robot control, particularly through his influential work on the Cassie robot platform. His 2018 and 2021 papers on deep reinforcement learning-based feedback control for bipedal locomotion — accumulating nearly 400 citations combined — helped establish RL as a viable alternative to classical model-based approaches, demonstrating that learned controllers could achieve robust, dynamic movement without relying on simplified dynamics assumptions. Berseth has since expanded his research scope considerably, contributing to landmark collaborative efforts in generalizable robot learning. His involvement in the Open X-Embodiment project and the DROID large-scale manipulation dataset — together gathering over 225 citations — reflects a growing commitment to building foundation models for robotics through diverse, real-world data. He has also advanced legged robot capabilities into more dynamic and precise behaviors, including bipedal jumping and quadrupedal soccer shooting. His intrinsically motivated learning work through SMiRL demonstrates a parallel interest in emergent behavior and unsupervised agent objectives, making Berseth a versatile and impactful voice shaping the future of autonomous robotic systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2Feedback Control For Cassie With Deep Reinforcement Learning188 citations · 2018
- 3
- 4DROID: A Large-Scale In-The-Wild Robot Manipulation Dataset108 citations · 2024
- 5
- 6
- 7Robust and Versatile Bipedal Jumping Control through Reinforcement Learning55 citations · 2023
- 8Feedback Control For Cassie With Deep Reinforcement Learning21 citations · 2018
- 9
- 10SMiRL: Surprise Minimizing RL in Dynamic Environments17 citations · 2019