Tanmay Shankar

Meta (Israel), Indian Institute of Technology Guwahati

Papers

Total Citations

H-Index

About

Tanmay Shankar is a leading researcher in robot learning, with a focus on enabling robots to discover, compose, and sequence motor skills from demonstrations. His work lies at the intersection of hierarchical reinforcement learning, imitation learning, and unsupervised skill discovery. Shankar’s most cited paper, “Discovering Motor Programs by Recomposing Demonstrations” (2020, 22 citations), introduces a novel approach to automatically learn recomposable motor primitives from large-scale manipulation data, bypassing the need for manually defined primitives. In “Hierarchical Reinforcement Learning for Sequencing Behaviors” (2018, 19 citations), he advances the options framework to learn both low-level control policies and high-level sequencing strategies. His work on “Learning Robot Skills with Temporal Variational Inference” (2020, 13 citations) further demonstrates unsupervised discovery of robotic options from demonstrations. Shankar has also contributed to visual navigation and deep reinforcement learning with recurrent convolutional networks. His research has been published at top venues including ICLR and RSS, and his methods have significant implications for scalable, general-purpose robot learning. With over 67 total citations, Shankar is shaping the future of autonomous skill acquisition in robotics.