Papers
4
Total Citations
60
H-Index
3
About
Elliot Chane-Sane is a researcher advancing the frontiers of robot learning, with a focus on goal-conditioned reinforcement learning, constrained locomotion, and video-conditioned policy learning. His most influential work, "Goal-Conditioned Reinforcement Learning with Imagined Subgoals" (2021, 26 citations), introduces a method where agents generate and pursue imagined subgoals to solve temporally extended tasks, addressing a key limitation in standard goal-conditioned RL. In "CaT: Constraints as Terminations for Legged Locomotion Reinforcement Learning" (2024, 18 citations), Chane-Sane pioneers a framework that integrates hard constraints directly into RL for quadruped robots, enabling efficient and safe locomotion policies—a critical step for real-world deployment. His work "Learning Video-Conditioned Policies for Unseen Manipulation Tasks" (2023, 14 citations) empowers non-experts to specify robot goals via demonstration videos, pushing toward generalist agents that generalize to novel tasks. Most recently, in "TD-CD-MPPI" (2025), he tackles long-horizon constrained control with a novel temporal-difference approach. Chane-Sane’s contributions bridge RL theory and practical robotics, earning recognition for making complex tasks more accessible and constraint-aware.
Research Focus
Key Achievements
Top Papers
- 1Goal-Conditioned Reinforcement Learning with Imagined Subgoals26 citations · 2021
- 2
- 3Learning Video-Conditioned Policies for Unseen Manipulation Tasks14 citations · 2023
- 4