About Me

I am CS PhD student in robotics & machine learning at the Robotics & Embedded Systems Lab (RESL) under Dr. Gaurav Sukhatme. I have 5+ years of work experience, working primarily on robotics controls problems outside Boston. I have experience in mobile robotics (SLAM), system modelling, control theory, machine learning (especially reinforcement learning), and computer vision. See my résumé and academic CV for more information.

My research interests are geared towards making robots perform regular day-to-day activities in unstructured environments (e.g. cooking in homes). I am particularly interested in the application of reinforcement learning and hierarchical planning for robotic manipulation tasks.

Research Internship @ Bosch

This week concludes my fabulous internship at Bosch Research. Through this summer, I have met a lot of great researchers and roboticists. I learned a great deal about the intersection of RL and robotics, implementing classical controllers on real robots, setting up environments and infrastructure, sim2real transfer, and a host of other concepts.

I was also glad to be  invited to give a technical introduction on reinforcement learning to my research department!!

Talk on an Introduction to RL

I was invited to give a talk on the basics of reinforcement learning to my department at Bosch during my internship. See slides below.

Awarded Amazon Research Awards 2018

A proposal I contributed to (with Ryan Julian) for the Amazon Research Awards was recently awarded to our lab (PI: Dr. Gaurav Sukhatme)!

See full list of awardees here.

Inverse Reward Design

Gave a presentation to Machine Learning folks on Inverse Reward Design, starting with a (very) brief survey of Deep RL

Inverse Reinforcement Learning in Java (using BURLAP)

Inverse Reinforcement Learning for automated grading

This spring semester, I did research on using Inverse Reinforcement Learning to learn a skill and test users, thereby creating automated graders for testing these skills. This is applied to sensorimotor (physical) skills, and a proof of concept is shown for a toy problem of navigating a car in a parking lot. See the presentation and links below for more details.

Other useful links:
PDF of slides
Code: https://github.com/gautams3/IRL_IntelligentGrading