Ajay U. Mandlekar – Research
Recent News
[December 2020] We recently extended RoboTurk to enable human-in-the-loop teleoperation and developed Intervention Weighted Regression, a simple and effective algorithm to learn from such interventions. We showcase results on two challenging contact-rich manipulation tasks requiring precision to solve.
[December 2020] We recently extended RoboTurk to enable multi-user collaborative teleoperation on challenging multi-arm tasks. We collected datasets on novel two-arm and three-arm tasks and developed a residual policy framework to address the problem of learning from multi-user demonstration data.
[December 2020] Check out our new preprint on Deep Affordance Foresight, a new model that extends the classical definition of affordance to enable long-horizon planning.
[September 2020] We released robosuite v1.0, a modular simulation framework and benchmark for robot learning.
[July 2020] We are organizing the Advances and Challenges in Imitation Learning for Robotics workshop at RSS 2020.
Publications
 |
Human-in-the-Loop Imitation Learning using Remote Teleoperation |
 |
Learning Multi-Arm Manipulation Through Collaborative Teleoperation |
 |
Deep Affordance Foresight: Planning Through What Can Be Done in the Future |
 |
robosuite: A Modular Simulation Framework and Benchmark for Robot Learning |
 |
Learning to Generalize Across Long-Horizon Tasks from Human Demonstrations |
 |
IRIS: Implicit Reinforcement without Interaction at Scale for Learning Control from Offline Robot Manipulation Data |
 |
Controlling Assistive Robots with Learned Latent Actions |
 |
Scaling Robot Supervision to Hundreds of Hours with RoboTurk: Robotic Manipulation Dataset through Human Reasoning and Dexterity
|
 |
AC-Teach: A Bayesian Actor-Critic Method for Policy Learning with an Ensemble of Suboptimal Teachers |
 |
RoboTurk: A Crowdsourcing Platform for Robotic Skill Learning through Imitation |
 |
Adversarially Robust Policy Learning: Active Construction of Physically-Plausible Perturbations |
 |
Risk-sensitive Inverse Reinforcement Learning via Coherent Risk Models |