My research goal is to develop continually aligned and helpful robots using demonstrations, explicit and implicit feedback, and intrinsic motivation. Always open to talk about AI safety, robotics, and reinforcement learning. Feel free to reach out over mail.
Learning latent manipulation policies that can be transferred to robots of different morphologies
via cycle consistency and a latent dynamics constraint.
Simultaneous Localization and Mapping using IMU and Lidar data. Estimated trajectory is further used
to recreate a texture map of the floor from RGBD images.
Optimal Stopping Dwait Bhatt, Deevanshu Goyal
The
Googol Game
Pitting human intuition against an optimal solution to a variant of the Secretary Problem.