Michael Luo

Santa Clara University, University of California, Berkeley

Papers

Total Citations

244

H-Index

About

Michael Luo is a leading researcher in safe reinforcement learning and interactive imitation learning, tackling fundamental challenges in deploying autonomous systems in the real world. His most impactful contribution is **Recovery RL**, an algorithm that addresses the critical safety-exploration tradeoff in reinforcement learning. By leveraging offline data to learn recovery zones, Recovery RL enables robots to explore uncertain environments while maintaining safety constraints—a breakthrough that has garnered over 200 citations and is shaping the future of reliable autonomous systems. Luo also developed **LazyDAgger**, a novel interactive imitation learning framework that reduces the burden on human supervisors by minimizing context-switching during corrective interventions. This work improves the efficiency of human-robot collaboration, making it more practical for real-world training. His research has been recognized at top venues like the Conference on Robot Learning (CoRL) and the International Conference on Robotics and Automation (ICRA). With a clear focus on bridging the gap between theoretical RL and safe, deployable robotics, Luo’s work is essential reading for anyone interested in building intelligent systems that can learn and act without compromising safety.