Trevor Russo, Entry #1, Learning (with difficulty) about Point Clouds

            My first week at Carnegie Mellon University has been a mix of failure and adjustment. On the first day, I grabbed my bike, my laptop and some other supplies and rode over to CMU’s Newell-Simon Hall, home of the Robotics Institute. However, upon arrival I discovered that I was early by at least 2 hours, so I decided to read some papers that my grad student Venkat had sent to me. Of course, most of it was written in really complex robotics terms, so I made the best I could of it before I ventured down to my lab. After some time, I was joined by two fellow interns, Joe and Shivam, both college students. We went out to lunch, and after that I was ready to begin work. Venkat had given me the task of familiarizing myself in the Point Cloud Library, or PCL for short. This software allows for 3D object manipulation by breaking it up in to a collection of dots, or a point cloud. I worked for 6 hours reading overviews of the software, and then I attempted to create my own program through the tutorial. Unfortunately, the second challenge on the website was too hard for the initial go, so I had to learn how to use software called CMake, which creates crucial files that allow for said software to run. After a long day of asking many questions and progress, I headed home. 

 

My Workstation

The next few days were much more difficult than I expected. Being good at computers and knowing secret commands on Windows is one thing. However, the tutorials that I had been attempting got increasingly difficult, as I was asked to accomplish tasks that I had never encountered on a system that I have rarely used. The system that I use, known as Ubuntu, doesn’t run through applications. Instead, I have to type commands through a really complicated terminal system. Often, I would try and run the code that I had written or modified and instead of seeing the success indicator, some stupid error would pop up and I would groan in disgust. Other times, I would see something, not understand what it meant, and spend 2 hours learning about it so that I could. Coding is much more difficult than I had expected. To help smooth the process out though, Venkat has told me to write a summary of the PCL tutorials that I had been learning about so that he can check whether I am knowledgeable enough to proceed on the creation and manipulation of point clouds.

The AI Secretary in the Entrance of the Robotics Institute

In between this work, I’ve attended a couple of talks. The talks are incredibly complicated and thus difficult to understand the meaning behind them, but they have taught me a valuable lesson: that not all research is good research. The first talk I saw was presented by a British fellow at the Imperial College of London. The presenter talked about deep learning, which is when a machine uses data and trials in order to “learn” how to better accomplish a task. Unfortunately, his research applied it to picking up blocks with a robot arm, and in the middle of the talk one of the professors called him out for it. The professor explained that to throw out a process like deep learning for such an easy task was preposterous. Instead of breaking into a prestigious field of robotics by being one of the first to introduce deep learning to robotics, the man instead stood up there looking foolish.

Working with Point Clouds

Even though I haven't been able to explore next, Pittsburgh is still an incredible city. There's the Pirates, a beautiful park, 5 museums and of course lots of good food. So far, my favorite place to eat has been the Rose Tea Cafe, which is a takeout Taiwanese place about a minute's walk from my apartment. The beef and broccoli there is incredible, although I'm sure there's a catch to that somewhere. The nice thing about the city is that everything is within reach of a 5 minute bike ride, so getting to work and back is extremely easy. Now that I've settled in after my first week, I finally feel that my next week will be much smoother, both in coding and getting to do things around the city.