Papers
12
Total Citations
347
H-Index
7
About
Roy Fox is a leading researcher at the intersection of imitation learning, hierarchical reinforcement learning, and real-world robotics. His work addresses fundamental challenges in enabling robots to learn complex behaviors from human demonstrations while remaining robust to distributional shift. Fox is best known for introducing DART (2017, 78 citations), a noise injection framework that bridges the gap between off-policy behavior cloning and on-policy methods, significantly improving imitation learning robustness. He has also pioneered deep continuous options discovery (DDCO, 2017, 50 citations) and multi-level hierarchical skill discovery (2017, 70 citations), enabling robots to autonomously learn reusable, temporally abstract behaviors that accelerate reinforcement learning. In applied robotics, Fox has demonstrated impactful results in surgical automation—developing a two-phase calibration procedure for cable-driven robots performing autonomous debridement (2018, 77 citations)—and in home automation through multi-task hierarchical imitation learning (2019). His work on statistical data cleaning for demonstrations (2017) addresses the practical challenge of noisy human data. With over 340 total citations across his top papers, Fox’s contributions are shaping how robots learn efficiently and robustly from humans in both simulated and physical environments.
Research Focus
Key Achievements
Top Papers
- 1DART: Noise Injection for Robust Imitation Learning78 citations · 2017
- 2
- 3Multi-Level Discovery of Deep Options70 citations · 2017
- 4
- 5Multi-Task Hierarchical Imitation Learning for Home Automation24 citations · 2019
- 6
- 7Robustly Adjusting Indoor Drip Irrigation Emitters with the Toyota HSR Robot14 citations · 2018
- 8Modular Framework for Visuomotor Language Grounding5 citations · 2021
- 9
- 10