Robotics Researcher & Engineer
I am an MS student at the University of Washington, Seattle, specializing in Robotics and AI with a focus on Learning based controls and perception for robots. I am a Research Assistant at the Robot Learning Lab, Paul G. Allen School of Computer Science & Engineering, where I work under the guidance of Prof. Byron Boots on cutting-edge research in robot learning.
Prior to joining UW, I did my BTech in Mechanical Engineering with minors in Robotics from IIT Gandhinagar.
I am interested in leveraging learning from demonstrations and reinforcement learning (RL) to automate systems with complex, real-world dynamics. Humans exhibit remarkable efficiency and adaptability, even with limited information and experience, not only in controlling their own bodies but also in operating machines and tools. My focus is on formalizing these human capabilities and transferring them into data-driven pipelines for robotics.
I have pursued projects in a diverse range of areas including Imitation Learning, Model-based RL, SLAM-based navigation, Deep Learning, Systems Development, and planning for Autonomous Agents. I'm passionate about applying these fields to contribute to technological advancements and explore new opportunities in Industry.
UW RACER (DARPA Funded Off-Road Autonomy Challenge)
Research Assistant October 2023 - Present
Developed the pipeline for Robot Learning on a Mushr Car in IsaacLab, implementing Delta Sampling MPPI based rollouts as a high level planner, created pipeline for Real2Sim transfer with getting real elevation data and terrain semantics embedded in IsaacLab Environment. Currently exploring Inverse Reinforcement Learning (IRL) based methods for Learning Reward functions from expert demonstrations.
Demonstrated that off-road driving is an increasingly difficult dynamics modeling problem in the aggressive regime. Quantified aggressiveness on analytic (no-slip and slip) and learned dynamics models. Evaluated on simulated (BeamNG) and real (Washington) aggressive driving data.
Johns Hopkins University
Research Intern May 2022 - Feb 2023
Designed and manufactured a tactile sensor-equipped finger for surface palpation, incorporating Gaussian noise to simulate real-life faults. Engineered a novel technique using genetic algorithms to achieve fault tolerance in tactile sensors. By injecting faults into sensor arrays, the method identifies faulty nodes and generates non-faulty offspring through fitness calculations. The algorithm's effectiveness was validated by integrating the sensor array with a prosthetic hand.
Engineering Services International (Now NewAge Robots)
Research Software Engineer August 2021 - Jan 2022
As a key member of the software team, I contributed to building self-driving capabilities for a new product by integrating essential hardware such as the Jetson TX2, stereo cameras, and LIDAR. I designed a ROS wrapper to streamline autonomous robot operations and implemented the RRT* algorithm to enhance navigation speed. Additionally, I explored and applied the TEB local planner, optimizing the system for more efficient autonomous navigation.
AgroBot, IIT Kanpur
Research Engineering Intern May 2021 - Aug 2021
Studied various sensors, actuators, and drivers used in the Agro-Bot and conducted real-time simulations using URDF in Gazebo. Designed a comprehensive ROS wrapper, mapping out the connectivity between nodes and topics through a detailed flow chart. Developed and modified ROS packages in Python, gaining hands-on experience in integrating robotic components and enhancing system communication, which provided valuable insights into the design and operation of autonomous agricultural robots.