Thu 03 October 2013

Hello research world!

Posted by ankur in Research (861 words, approximately a 4 minute read)

It's been just about a year since I started my research career as a masters student here at the University of Technology, Sydney. I'm working on bio-mimetic navigation for this course, focussed around computational modelling of head direction cells. This post is a traditional "Hello world!" post and documents my work in general.

As expected, a research course is completely different from any other. Firstly, I have no classes to attend, no assignments to work on, no exams to study for.This doesn't imply that I haven't any work. A year into my course, I think a structured coursework degree is much easier to negotiate than a self structured research degree. For instance, in a coursework degree, you know exactly what to study, you have a fixed schedule to follow, you know exactly how you'll be graded. In a research degree, however, you need to find your own material to study, you need to plan your own schedule, you need to have results that you think will get you a good grade. While students run off on holiday after their exams, I still go back to work every day! The most difficult part, perhaps, is the amount of self discipline that you need to possess. I regularly struggle with this: waking up early, following a fixed schedule, sleeping on time, following my day plan; aren't really easy to carry out when you don't have deadlines to scare/motivate you. As a result, I spend quite a lot of time in introspection (refer to my post on journal writing).

An introduction to my research

I work in an area called "bio-mimetic navigation". As the name suggests, it's simply about mimicking the various navigational systems found in nature, and if possible applying them to robotics. Modelling biological navigation can be done at different levels, though. I work on a neuronal level with **head direction cells**. These are specific neurons that have been discovered in the rat hippocampus that maintain information about the rat's heading. Of course, I can't go into the details of the system that have been discovered over the three decades of research here. The wikipedia page is a good place to start for any one interested.

Why bio-mimetic navigation?

When I started off with my course, my research topic was much broader: "mobile robot navigation in a dynamic environment". It was designed to be broad enough to give me the opportunity to review literature and decide for a focus area by myself. Like everyone, I began with classical robotics. I read on SLAM, inertial navigation, sensor fusion and learnt quite a bit about the navigation systems that robots utilize, the concepts that make up these systems. A few weeks into my literature review though, I tried to compare these methods to the way I navigate personally. I found that I hardly used any of the classical robotic techniques. "Why then, should robots use them?", I wondered. Curious, I began reading on animal navigation, and ran into bio-mimetic navigation. I learnt how animal navigation is very very different from classical robotic navigation techniques. I now belong to a school of researchers that are interested in developing navigation techniques from nature, and in comparing these with classical robotic techniques. I'm very new to the field though, but after this two year masters course, and a Ph.D. to follow, I hope to be right in the thick of things.

My work with head direction cells makes my research multidisciplinary. I need to know a little about neuroscience, along with my computer science basics. Since I'm a computer science graduate, my knowledge of neuroscience isn't as good as I'd like it to be, but it's getting better every week.


The goal attached to my masters course is to prepare me for a Ph.D. degree. While I did know that I wanted to work with robotic navigation, with neural networks, an undergraduate degree didn't give me enough depth of knowledge to make an informed decision on the exact research problem that I wanted to work on. The masters course, therefore, gives me much needed time to read my literature, decide my stream and develop skills for a career in research. While my masters plan is closely aligned to my goal, I'm also going to try and make some contribution to the field, however small. (It's always good to have a few articles published when you're a new researcher looking for opportunities to grab.)

At present, I've read many *many* papers on head direction cells and their modelling. I'm currently working on extending one of these models, my contribution to the field. I was recently assessed, and my assessors decided my work was good, and that I could carry on my course here for the remainder of the course. If you're interested, you can read view my assessment presentation and my report here. My assessment report contains quite a comprehensive literature review. The reference list will be quite helpful to people interested in the field. (If you do pick any thing from my report, please do cite it).

I'm now engaged in making some contribution to the field. I'll post again when there's something worth reporting. Cheers!