Aravind Rajeswaran
University of Washington, University of California, Berkeley
Papers
18
Total Citations
584
H-Index
9
About
Aravind Rajeswaran is a leading researcher at the intersection of robotics, computer vision, and artificial intelligence, with a primary focus on enabling dexterous manipulation and general-purpose robot learning. His seminal work on dexterous manipulation with deep reinforcement learning (RL) has been highly influential, demonstrating that complex, multi-fingered hand control can be learned efficiently and at low cost—a breakthrough that has garnered over 170 citations. He pioneered the use of demonstrations to bootstrap RL for high-dimensional control tasks, and his research on visual representations, notably R3M, has shown how pre-training on diverse human video data (like Ego4D) can dramatically improve sample efficiency for downstream robotic tasks. Rajeswaran has also made key contributions to offline RL, embodied question answering (OpenEQA), and scalable imitation learning (CACTI). His work on the "Train Offline, Test Online" benchmark addresses critical reproducibility and generalization challenges in robot learning. With over 500 total citations across his top papers, Rajeswaran is recognized for advancing the frontier of learning-based control for real-world robotic systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3R3M: A Universal Visual Representation for Robot Manipulation82 citations · 2022
- 4OpenEQA: Embodied Question Answering in the Era of Foundation Models46 citations · 2024
- 5Learning Deep Visuomotor Policies for Dexterous Hand Manipulation43 citations · 2019
- 6
- 7
- 8Train Offline, Test Online: A Real Robot Learning Benchmark16 citations · 2023
- 9Offline Reinforcement Learning from Images with Latent Space Models16 citations · 2020
- 10