Motivation

Mobile Data Management has always been my key focus. Initially I was looking at application architectures comprising of end-to-end systems and proxy-based systems. Work in this field did interest me a lot, but I was not able to think of something really novel in this area, something that has not been done and which is practically useful.

During my Mobile Computing Class, I had to give a presentation on Sensor Networks and that was when I got really interested in this area. Most of the computing today is centered around providing services to humans, Sensor Networks propose a paradigm shift in terms of moving computing towards our environments. Research in Sensor Networks is relatively new (lets say, about 3 year old history) and this made the area much more exciting to investigate. There is a great deal of uncertainty existing today as regards Sensor Networks improving our day to day lives, but I am confident that only a few years down the line, sensor networks will be everywhere, not just the Military Environments that they cater to right now.

I decided to pursue work for my thesis in this area. First of all, data management served as one of the key focal points of research in this field and this directly complied with my long term goal. Thereafter, having to explore a relatively new idea is much more easier than something that has been investigated for decades (like operating systems, databases etc). During my summer internship at Hewlett Packard Lab, I had a chance to work with CoolTownSensing Infrastructures. This exposure further propelled my interest in sensor networks. There are three key aspects that I want to gain from my thesis.

1. Novel Idea: Generating an Idea that is different from the rest of the work is the most difficult part of any thesis. My key effort is to brainstorm myself and come up with something really different, something new and something that makes practical sense.
2. System Level Implementation: I have worked with Java for quite sometime now, but this time, I wanted to explore the Systems aspects of implementation software. Particularly, I am interested in using network simulators (like ns-2, sensorsim) for experimenting with my ideas and evaluating them
3. Engineer an Experimental Testbed: I find real challenge in carefully engineering a testbed for evaluation of schemes and algorithms that we develop. To measure a few parameters and keeping all others constant is a really difficult job and I want to master this once I get done with my thesis.

Areas of Interest

I am interested in two key research areas in sensor networks.
Having done a literature review of the current state of research, I find that most of the work hovers around with energy as the prime consideration. Routing techniques, data propagation techniques, hardware design, all have to center their attention on energy conservation. This propelled me to think about data propagation scheme(s) with energy conservation in mind. There are a few techniques already in place like SPIN, Directed Diffusion, Leach etc, but all these techniques do have some underlying assumption, which when relaxed, might make the performance degrade. So I am thinking of devising a new scheme or modifying existing ones based on the domain of interest, that being Smart Spaces(Smart Rooms, Smart Offices etc). The exact idea is still under scrutiny and will be published on the web, once papers have been rolled out on the same.

Thereafter I started thinking about possible applications of Sensor Networks in SmartSpaces. Here the energy is not the be all and end all of things. There are other issues that need more attention then energy conservation. I am still not yet quite convinced with the idea that sensor networks can be applied to smart spaces, but the investigation goes on.

Now I am in my final phase of my Masters Thesis. The second phase involved a lot of reading into new research directions in sensor networks and interesting work was that at Cornell wherein they view Sensor Network as a Distributed Database. The stress now is on query processing in such networks and for sensor networks of considerable size, huge data streams would be generated. The key question would be ways to handle such streams for extracting the relevant data in a timely fashion and in an efficient manner. During this last phase, I am trying to give direction to my MagicWeaver Simulation framework. As of now its a general purpose simulator, but then we want to make it focussed towards particular types of applications. To this end, we are looking at providing support for Streaming Applications and compare the performance of MagicWeaver with existing simulators like SensorSim.

Proposed/Accomplished Work

As a first step towards Sensor Networks, I carried out an exhaustive literature search of data management research in sensor networks. Through this I came across a fair bit of work in desiging new algorithms for data propagation in sensor networks.

My Second step was to implement a system on sensor network to gain a better understanding of the domain. To this end, I build MagicWeaver: An Agent based Simulation Framework for Wireless Sensor Networks. This experience gave me a deeper understanding of the issues prevailing in designing data management techniques for sensor networks and made me understand, SPIN and Directed Diffusion Protocol really well.

MagicWeaver is a Simulation Framework for wireless sensor networks. It is a Java/JADE based Multi-agent system modelling sensor networks and its operation. You can have a look at the Software API here. The heterogeniety of algorithms that are prevalent in this domain, warrants some kind of framework that has the flexibility to incorporate these and present results. It provides several abstract models that are pertinent in sensor networks and provides a runtime environment for incorporating user-defined instances of these models. It provides graphical utilities for monitoring the network state and the data that is received at the base station. For further details of MagicWeaver,it's design, source code, please email me .

To complete this last phase, I am working in providing support for streaming applications in MagicWeaver. A paper is planned to be published in this area, please check back later, once it has been accepted.

For the details on the first paper, please click here