Carnegie Melon University- Nithin Vasisth

923 Views, Posted on: July 17, 2016

The content on this website is strictly the property of Insight and the Students’ Gymkhana IIT Bombay. If you wish to reproduce any content herein, please contact us:
Chief Editors: Shreerang Javadekar, Shreeyesh Menon
Mail to: insight@iitb.ac.in

Nirhin Vasisth is a 3rd year Undergraduate pursuing a B.Tech in Mechanical Engineering.

About two minutes of stunned silence, followed by going totally berserk and running around not knowing what to do was my reaction when Aditya Kale, after teasing around, eventually told me “Potty … Congrats … Tera ho gaya CMU me”.

The Robotics Institute, Carnegie Mellon. This was not expected and needless to say, this meant a lot.

null
The Field-Robotics-Centre Highbay at The Robotics Institute

In this blog I’ll be going over some aspects of the work I did at my internship, and in the last few sections I’ll let you know ki maine US me kya-kya machaya (:P)

Pre Intern

This internship came in through PT Cell in the form of a 4-week assignment challenge. The purpose of the assignment was to familiarize us with the basic concepts of Motion Planning, Perception and/or Semantic Segmentation using Deep Learning. These may sound fancy, but the assignment takes you through the basics in a systematic way, and all one has to do is stick with it.

This is a standard selection procedure at The Airlabs and each year they modify the assignment to suit their current research needs. This year’s assignment can be found here.

After the selection, I immediately contacted my mentor for additional papers and details about the work. I was not going to let this summer go to waste. (I got re-allocated after I reached. So much for enthu :P)

Intern Work

At The Airlabs, I was a part of the AACUS project, which is responsible for the development of advanced autonomous capabilities for unmanned air systems ( A freaking autonomous helicopter! )
I was given terabytes of flight data, and was asked to come up with a cost function that resembles what the pilot is subconsciously trying to optimise. Everything was very new to me and God, was I thrilled to be there!

My work was related to the field of Imitation Learning/ Inverse Reinforcement Learning, where you have a set of example behaviors and you have to come up with a mathematical formulation to represent that behaviour. (It may be through a cost function, an optimal policy based on state features, etc.). If you want to know more about this field of research, feel free to get in touch with me.

I spent much of my internship adding functionalities to a tool that generates reports from the flight data and developing tools for data analysis and representation. So, by the time I got to the actual research part, I barely had any time left and I ended up working on a small subset of the problem.

Other Labs at CMU

Other than my project, I visited a lot of labs at random and got to see some really cool stuff:

  • Gates Highbay: Lunar Rover prototypes, Agile quadcopters working in swarms
  • CMU’s Hyperloop prototype. (A small working model for the hyperloop competition)
  • PR2, Jaco Robotic arm in action, working on basic manipulation stuff
  • Biped Walking robots
  • Soft Robotics Lab: Artificial skin sensors, muscles and soft robots
  • Biorobotics lab: Snake robots and surgical robots (The Da Vinci prototypes)

People were always helpful and took out the time to explain what they were working on, and it was very humbling to be around.

Final Presentation

It was a disaster.

*shaking my head*

Every intern at Airlabs is required to present his work in a formal talk, and typically a mail is sent to everyone at The Robotics Institute. Owing to the fact it was Monday morning during summer, about 7 people showed up.

The mistake(s) I committed are:

  1. I didn’t have a good ppt ready, let alone have it reviewed by the professor
  2. Didn’t give out a practise talk in front of my lab group
  3. No pizzas && Late Notification == less people

Because I needed to show at least some good results on flight data, I ended up not sleeping the previous night and working on a new method (This worked).
Although, the presentation was not ready and I thought I’d wing it. (This did not work).
The result was that I was still typing into my ppt as people started to come in and I was extremely nervous. There was even a moment when I just stopped talking and stared at the audience for a while.

*still shaking my head* *never again*

My professor gave me some feedback and I’ll share it here, in case it might be useful to some of you.

“It would have been great to see more details on the precise problem and what exactly your approach is. What is the input data? What is your problem setup and so on. This template might be useful to structure your future presentations and/or write-ups.
– Context: What is the context of your work, what is the start of the art
– Need: What is the lack of the start of the art
– Task: What is the task you want to address in your work (should be 1 sentence)
– Object: How do you plan to address your task?
– Expected results: What is your hypothesis on the outcome and how will you measure it?
– Conclusions: What are your expected conclusions
– Perspectives: How could you further the proposed work.“

Pittsburgh

Pittsburgh is a lovely place to live. Really calm, peaceful and quiet. Except when it’s not. Sporting events are the time when the whole of downtown comes alive and you can hear the entire streets filled with cars honking and people shouting “Let’s go Pens!” in unison. The city takes its sports very seriously and it is really fun to be around.

I had my accommodation arranged with a few other interns at CMU. From cooking completely arbitrary meals, late night walks to the US-mein-hai-toh-strip-club-chalte-hai talks, it has been really fun.

Moms make a big deal out of cooking. None of us had cooked before, and it was really exciting to cook our own food. We started out great: Cooking our meals everyday, doing the dishes regularly, cleaning the house, and even folding our clothes… until a few days later…

null
…. we started living like pigs. (We really miss our moms)

Pro tip #68: Undercooked or overcooked. Everything tastes good with garam masala, mayo and aloo bhujia.

Outside Pittsburgh

Chicago | Peoria : My relatives were nice enough to have me over a weekend; and we went around Chicago and it was beautiful. The Sears Tower, a boat ride around the Chicago skyline and that black building in The Dark Knight. The city had a very dreamy feel to it, it was like walking into a movie (or maybe it was sleep deprivation). And I got to eat ghar-ka-khana, that I had started missing so dearly.

New York : I love New York! This city totally lives up to its hype. Even after days of roaming around, all of us were coming back with “Shit! Kitna kuch baaki reh gaya!” feels. And the whole trip had a very grown-up feel to it: meeting your friends to hang out at NYC! and it would’ve been quite boring without them.

null
null
Washington DC: …hmm.

Summing Up

Needless to say, this was an amazing experience. Wouldn’t have been the same without these guys and looking at these pics someday is guaranteed to make me super-nostalgic.

null

As for the project, I’ll be continuing my intern work as my final year project. And I’m really looking forward to this next phase of my life 😀

Thanks for reading.

Leave a Reply