Teaching

teaching

I wish to empower students to solve real-world robotics problems. Our joint objective is to develop simple and portable concepts, apply them to messy real-world problems, and mine new insights from the process. Many of the concepts I carry around today have been shaped by thoughtful students, colloborators, teachers and thinkers.

CSE 490R: Mobile Robots

https://courses.cs.washington.edu/courses/cse490r/19sp/

This was the first undergraduate course that I taught from scratch. My goal was to enable students to walk in with little to no knowledge of a robot and, in 11 weeks, walk out with a clear understanding of principles and challenges for any robotic system. With this in mind, we made brand new lectures that covered the principles of inference from sensor data, local control, global planning and robot learning. We designed a new set of 13 racecars, created programming assignments, tutorials aimed at guiding students through building their own fully autonomous robot. Every team successfully competed the final demo where they autonomously raced around 6th floor of CSE collecting hidden "gold". I am grateful to the superhuman TAs who were really the backbone of this class. As a first edition, the course was very well received (4.8/5.0) and the lectures much appreciated (4.9/5.0). While the next edition will be bigger and better, this quote was very encouraging:

... is an amazing lecturer who presented stimulating concepts each week and constantly challenged us to frame the material in novel ways and think like actual robotics researchers. TAs put incredible effort into making applicable assignments/labs, and though they were dense and a bit overwhelming at times, the labs vastly improved my skills and knowledge in so many ways!

.

mushr

This entire effort is open-sourced. For more details visit: https://prl-mushr.github.io/.

Online Lectures

I recently started putting out a series of online lectures. They cover a range of topics, starting with my favourite: imitation learning! I had a lot of fun experimenting with different mediums and plan to put up more.

Imitation Learning: A Series of Deep Dives

In this 10-part series, I dive deep into imitation learning, and build up a general framework. A journey through feedback, interventions and more!

Core Concepts in Robotics

A series that revisits core concepts in robotics in a contemporary light.

Advising

I have had the privilege of working with an incredible group of students and collaborators over the years. The happiest moments of my research life have certainly been during deep, insightful discussions with students. To find out more, feel free to reach out to any of them below!

PhD Students
Gokul Swamy, CMU 2020 - Current
Aditya Mandalika, UW 2018 - Current
Gilwoo Lee, UW 2018 - 2020
Brian Hou, UW 2018 - Current
Liyiming Ke, UW 2018 - Current
Matthew Schmittle, UW 2018 - 2019
Mohak Bhardwaj, GaTech 2019 - Current
Jonathan Spencer, Princeton 2019 - Current
Brad Saund, UMich 2019
Rogerio Bonatti, CMU 2016 - 2020
MS Students, Researchers
Swapnil Asawa, Aurora 2021 - Current
Johan Michalove, UW 2018 - 2019
Ajinkya Kamat, UW 2018 - 2019
Rosario Scalise, UW 2018 - 2019
Vishal Dugar, CMU 2015 - 2017
Mohak Bhardwaj, CMU 2015 - 2017
Undergraduate Students, Researchers
Matt Cheung, Aurora 2020 - Current
Andrew Pham, Aurora 2020 - Current
A. J. Kruse, UW 2019
Max Thompson, UW 2019
Rajat Jenamani, IIT 2019
Rahul Vernwal, IIT 2018
Gleb Shevchuk, Stanford 2018

Workshops

Workshops are easily the most fun part of a conference! A good workshop is a long form discussion of a timely topic and is much appreciated by newcomers to a field. I have found this to be the maximally impactful activity for the robotics community and continue to seek co-organizers to explore new topics. Here are some of the workshops I have organized:

Machine Learning in Robot Motion Planning, IROS 2018

https://personalrobotics.cs.washington.edu/workshops/mlmp2018/

This was the most enjoyable workshop that I have organized. The topic, which I am quite passionate about, was indeed timely - how do we formalize priors in motion planning? Where do we get these priors from? What can we solve that is currently unsolved? The talks and papers were excellent and the discussions went well over time. The most satisfying part was that the local students from University of Madrid really appreciated the workshop and the exposure to such discussions.

Imitation Learning and its Challenges in Robotics, NeurIPS 2018

https://sites.google.com/view/nips18-ilr/

NeurIPS was a perfect platform for this workshop which had one central message - think about humans! A lot of imitation algorithms forget that humans provide the labels and these humans get tired, are often inconsistent, have trust issues and refuse to crash into things! We had a lot of good discussions that continues to influence my research today.

Complex Collaborative Systems: Closing the Loop, Learning, and Self-Confidence

http://theairlab.org/iros2017_workshop/

This was my first workshop as an organizer. It was co-organized by UTRC and we were fortunate to have a heavy industry presence. As a result, we had a good discussion on practical issues when we close the loop on complex systems that employ learning. The discussion went to interesting places like Not all data is information, Many real-world problems are satisficing, not reward maximization and Self-contained autonomy is a myth.