Papers
30
Total Citations
979
H-Index
12
About
Ajay Mandlekar is a robotics researcher whose work sits at the intersection of robot learning, imitation learning, and simulation infrastructure. His research addresses one of the central challenges in robotics: enabling robots to acquire complex manipulation skills efficiently and safely, without the prohibitive sample costs of traditional reinforcement learning. Mandlekar is perhaps best known for developing scalable data collection and learning infrastructure. His RoboTurk platform pioneered crowdsourced robot teleoperation, enabling the collection of large human demonstration datasets that were previously infeasible—work that has accumulated over 80 citations and directly influenced subsequent large-scale efforts. Building on this foundation, he contributed to Orbit, a unified NVIDIA Isaac-powered simulation framework for robot learning (226 citations), and played a role in the ambitious Open X-Embodiment collaboration (119 citations), which consolidated diverse robotic datasets to train generalist models. His research on imitation learning generalization—particularly GTI for long-horizon tasks and his rigorous benchmarking study on offline learning—has provided both algorithmic advances and practical guidance widely adopted by the community. His earlier work on adversarially robust policy learning (128 citations) demonstrated prescient concern for deployment reliability. Together, Mandlekar's contributions have meaningfully advanced robotics toward scalable, generalizable, and practically deployable learning systems.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7Deep Affordance Foresight: Planning Through What Can Be Done in the Future49 citations · 2021
- 8
- 9RoboCasa: Large-Scale Simulation of Household Tasks for Generalist Robots27 citations · 2024
- 10