Sanjiban Choudhury

sanjibanc at cornell dot edu

prof_pic.jpg

Assistant Professor

Cornell Computer Science

I am an Assistant Professor at Cornell Computer Science where I lead the People and Robot Teaching and Learning (PoRTaL) group.

My research aims to build robots that work seamlessly alongside human partners in the wild. To this end, my work focuses on imitation learning, decision making and human robot interaction. I am interested in domains where a robot continually interacts with humans such as collaborative mobile manipulation and self-driving.

I work part-time at Aurora, where we use machine learning to enable safe, human-like driving. I did my Postdoc at University of Washington and Ph.D. in Robotics at Carnegie Mellon University. Much of my research has been deployed on real-world robot systems — full-scale helicopters, self-driving cars, and mobile manipulators.

Propsective students and undergraduate researchers see this note.

Selected Papers


2021_moment_matching

Of Moments and Matching: Trade-offs and Treatments in Imitation Learning
Gokul Swamy, Sanjiban Choudhury, Zhiwei Steven Wu, and J Andrew Bagnell
International Conference on Machine Learning (ICML), 2021
project page / paper / video / code

All of imitation learning can be reduced to a game between a learner (generator) and a value function (discriminator) where the payoff is the performance difference between learner and expert.


2019_eil

Learning from Interventions: Human-robot interaction as both explicit and implicit feedback
Jonathan Spencer, Sanjiban Choudhury, Matt Barnes and Siddhartha Srinivasa
Robotics: Science and Systems (RSS), 2020
paper / talk

How can we learn from human interventions? Every intervention reveals some information about expert's implicit value function. Infer this function and optimize it.


2018_data_driven_planning

Data-driven Planning via Imitation Learning
Sanjiban Choudhury, Mohak Bhardwaj, Sankalp Arora, Ashish Kapoor, Gireeja Ranade, Sebastian Scherer, Debadeepta Dey
The International Journal of Robotics Research (IJRR), 2018
Finalist for Best Paper of the Year
paper

Train planners (that operate on partial information) to imitate clairvoyant planners (that have full information) to choose optimal planning decisions. (applies to heuristic search, exploration planning, etc)

News

Nov 20, 2022 Looking forward to attending NeurIPS in person. Come checkout our two papers - Sequence Model Imitation Learning with Unobserved Contexts and Minimax Optimal Online Imitation Learning via Replay Estimation!
Oct 7, 2022 Giving a talk at Cornell AI Seminar on Imitating Experts with Privileged Information
Aug 1, 2022 Teaching a brand new grad course, Learning for Robot Decision Making. Check out the course trailer!
Jul 1, 2022 Excited to start my new group, PoRTAL, at Cornell!
Jun 1, 2022 Congrats to Gokul! Our paper on Causal imitation learning was selected for Long talk at ICML 2022. Also come check out our ICML paper on superhuman imitation learning with Brian Ziebart and gang from Aurora.
Apr 25, 2022 Excited to share a series of talks on Interactive Imitation Learning that I gave at Stanford Robotics Seminar, USC CS Colloqium, UW Robotics Colloqium and UPenn F1Tenth. Covers some of my current work at Aurora and future research that my lab at Cornell will be focusing on!
Jan 2, 2022 I recently started putting out a series of online lectures. Check out Imitation Learning: A Series of Deep Dives, Core Concepts in Robotics and Interactive Online Learning!
Oct 21, 2021 Excited to share the cutting edge self-driving tech behind the Aurora Driver! We use machine learning to make better driving decisions. I talk about how our learned decision making and interactive forecasting approach enables safe, human-like driving. Also presenting this at BARS 2021