Benjamin Swanson
Papers
4
Total Citations
93
H-Index
4
About
Benjamin Swanson is a leading researcher at the intersection of robotics and reinforcement learning, with a primary focus on enabling robots to continuously adapt and generalize across diverse tasks. His major contributions center on overcoming the limitations of fixed-policy robotic systems, demonstrating through works like "Never Stop Learning" (41 citations) and "Efficient Adaptation for End-to-End Vision-Based Robotic Manipulation" (13 citations) that fine-tuning and continuous adaptation are not just theoretical promises but practical, scalable solutions. Swanson’s work on MT-Opt (35 citations) and its scaling counterpart (4 citations) has been pivotal in advancing multi-task robotic reinforcement learning at scale, showing how general-purpose robots can efficiently master a wide repertoire of skills without prohibitive training times. His research has been instrumental in moving the field toward deployable, adaptive robotic systems that learn from their mistakes in real-world environments. With a growing citation impact and a clear trajectory toward practical, generalist robotics, Swanson is a key voice in the push for robots that truly never stop learning.
Research Focus
Key Achievements
Top Papers
- 1
- 2MT-Opt: Continuous Multi-Task Robotic Reinforcement Learning at Scale35 citations · 2021
- 3Efficient Adaptation for End-to-End Vision-Based Robotic Manipulation13 citations · 2020
- 4Scaling Up Multi-Task Robotic Reinforcement Learning4 citations · 2021