Sumeet Singh


Sumeet Singh 

I am a Research Scientist at Google Brain Robotics in NYC. I recently completed my Ph.D. in the Autonomous Systems Lab in the Aeronautics and Astronautics Department at Stanford.


PhD Defense 

I recently passed my PhD defense. Click on photo for video link. Additional quadrotor videos can be found here.

Research Overview

Broadly, I design algorithms for robot planning and control by combining tools from nonlinear control theory and reinforcement learning. Specifically, I am interested in (i) how to use control theory to stabilize and make robust, in a learning and operational sense, model-based reinforcement learning (RL), (ii) how to merge model-based and model-free RL algorithms in such a way as to alleviate issues such as high sample complexity and bias-induced errors, and (iii) rigorously evaluating novel algorithms within challenging experimental domains including aerial & wheeled ground vehicles, dexterous manipulation, and agile locomotion.

Some topics that I studied during my Ph.D. are listed below:

  • Robust motion planning for constrained nonlinear systems: Design of nonlinear control and planning algorithms for online generation of robust motion plans with guaranteed margins of safety for constrained robotic systems in cluttered environments (ICRA, WAFR, IJRR Preprint).

  • Control-Theoretic model-based Reinforcement Learning: Infusing model-based reinforcement learning algorithms with concepts drawn from nonlinear control theory, such as stabilizability and robustness (WAFR, IJRR Preprint).

  • Risk-sensitive inference and decision-making: Development and analysis of flexible risk-sensitive human behavior models and the design of stochastic control algorithms leveraging these models within a risk-sensitive dynamic game framework (IJRR, TAC, RSS).


Invited Talks