Hello, everyone! My name is Hiya, and I am a third-year undergraduate in the Department of Mechanical Engineering. My academic interests are in Systems and Controls, especially with applications in robotics and automation. Apart from that, I love to write, sing, and critically analyse period dramas and dystopian movies.
I was first introduced to control systems through my student team, UMIC, although I was then in the mechanical subsystem. As I explored it more from a theoretical aspect by taking a Minor in SysCon, I realised I enjoyed the combination of mechanics and math that the field provided.
As my interest in the topic increased, I decided to do a project under Prof. Ravi Banavar of the SysCon department. I had a fair idea of what subfield excited me the most in Controls. This only got consolidated as I spent many months learning the basics of control theory under Prof. Banavar.
Towards the end of the Second Year, Prof. Banavar and I finalised a problem statement: building a non-linear observer for Visual Simultaneous Localisation and Mapping. The project was in collaboration with the Systems Theory and Robotics Group at the Australian National University. It involved a drone that had to map the environment landmarks while simultaneously localising itself with respect to these landmarks. It must do this with a camera, a sensor that cannot measure depth but can only get the direction it sees the landmark. The objective of the observer is to convert these directions into distances. The problem is a little more complicated than explained here, but the ultimate aim was to implement VSLAM significantly lower cost.
During the summer, I was given a subtask: implementing an Equivariant Filter for a SLAM problem and tweaking its parameters. What is an Equivariant Filter? It is similar (not entirely) to a Kalman filter but is defined on an equivariant system. The objective was to find the best parameter (or what we call ‘origin’) that would make the error converge to zero the fastest.
As I explained before, the application I had to work on was VSLAM / SLAM. But that wasn’t my primary motivation to take this project. The math involved was differential geometry, along with some control theory concepts. I had to work on a system that was on a manifold, a Lie group in my case. It was observed that working on a system defined on a particular Lie Group reduced the computation cost compared to using something like Euler angles or cartesian coordinates. This, coupled with the fact that it was very new research (about a year old) and had much yet to be explored, was an exciting enough reason to pursue the project.
Before starting my project during the summer, I had some basic expectations- getting good results, learning a few new topics, etc. I thought I would be spending some months clearing some checkpoints set by the professor. However, it turned out to be so much more than that. In most research projects, the professors have a fair idea of what results to expect. However, in frontier research, it could go in absolutely any direction, and all you could do was keep rebuilding your hypotheses. I had no checkpoints, it was an open field, and I could run anywhere I wanted.
It was exhilarating but also gruelling work. We spent weeks trying different theories to corroborate the results we obtained and even had to scratch out most of them. Apart from these theoretical roadblocks, I had to figure out how to implement a filter, debug it and run simulations (which sometimes took hours!). I also had to familiarise myself with differential geometry, which was a little difficult considering that I haven’t had any formal introduction to it in my courses.
The work from home experience was a little constrained but not too much of a barrier. Most of my work was theoretical and could be done easily in an online setting, thanks to my tablet and stylus. We would meet once a week and discuss ideas and the progress of the project. I maintained a GitHub repository for all my reports and codes, making them easily accessible. I was also invited to a biweekly session where I could see the projects that other undergraduates were working on, which was excellent exposure for me (as I did not understand most of what they explained :P).
The highlight of the experience was that I could contribute fresh ideas because the topic was new for everyone involved. Some ideas worked, some didn’t. Each step helped me structure my thinking process in a particular way that would surely help me in the future, should I decide to pursue research. For example, I had an idea which I was able to back quite strongly. I was encouraged to look into it, but after some study, it turned out that it wouldn’t help us much at the end of the day. However, this idea got us thinking in another very promising direction, which ended up giving us good results!
The entire experience taught me to think critically, keep trying, and never give up. I realised, a seemingly bad result is also, in some sense, a “result” and holds much value in the research community.
Message to readers
To those interested in exploring research- getting in touch with a professor in the institute is a considerable advantage. Many are very active in their respective research and have a good idea of how to induct students into their work. I am incredibly grateful to have had a guide who gave me exposure in the field, even something as niche as VSLAM.
Trying out a project in the second year to understand what research entails is an excellent idea to figure out if it is something for you.
To those who are apping, institute professors are well connected to professors from other universities and can connect you with them. Being recommended by a guide professor from the institute would, at the least, increase the chance to get a reply from universities/professors.
My inclinations being primarily towards theoretical work, I was able to appreciate the experimental aspects more after my experience in the summer. I had obtained skills like debugging (which is huge for me, not being a coder), brainstorming, and developing an organised procedure to build any hypothesis. I had picked up topics like differential geometry and group theory, in which I had no background. But most importantly, it was just a lot of fun 🙂