The following figure shows my typical research methodology for projects like road traffic monitoring or human mobility
measurements. There is a component of (i) sensing or gathering data on the phenomenon of interest,
(ii) analysis of that data to gather insights, (iii) communication among sensors for coordination, and to the
Cloud for transferring data and/or inferences and (iv) deployment of system prototypes, that implement
the first three components, for long term evaluation. Communications can also be encrypted, if the application
needs privacy.
Since what sensor or analysis methods will work in a given scenario is unknown, I typically use an iterative
designing, experimentation and re-designing process. Projects start with offline analysis, where candidate
methods are evaluated for accuracy and efficiency. The best method identified is implemented into
a working prototype to do in-situ analysis, for demonstration and deployment. Finally, long term insights
are gathered from deployments, for temporal analysis of the phenomenon under consideration.
Thus I need three kinds of collaborators in my projects -- (i) technical experts: since I combine techniques from different areas of Computer Science (CS), like embedded systems, mobile computing, wireless networks, computer vision, applied machine learning, deep learning, operating systems and cryptography, I need experts to explain my necessity and learn from them the relevant topics in a particular area of CS, (ii) domain experts: since the problems I work on are at the intersection of Information Technology and society, I need help from goverment and private organizations to deploy my prototypes, like traffic control authorities for deploying my road sensors or shopping mall and airport authorities for human mobility measurements, (iii) co-workers: a typical system project is large and working as a team helps in faster progress. All my projects have been in teams with undergradute, Masters and PhD students and interns, where I have contributed 50-95% in terms of actual hands-on work.
[1] Road traffic measurement in developing regions
In my PhD, I collaborated with M.Tech students Prashima Sharma, Abhinav Maurya and Swaroop Roy, IITB staff Pankaj Siriah and Pramod Mhaske, and Research Engineers Swanand Kulkarni, Rupesh Mehta and Ramkrishnan Kalyanaraman. Prof. Bhaskaran Raman, Prof. Preeti Rao and Prof. Saketha Nath from IIT Bombay, and Prof. Amarjeet Singh from IIIT Delhi advised us at different stages of this work. This is an essay on how I built the collaboration with Bengaluru traffic police, with help from an Indian startup Mapunity for long term deployment and testing of my congestion sensors. Nascent Technologies at Ahmedabad, India has taken my RF based solution and built commercial prototypes for Gujarat traffic.
I also worked in this area as an intern at Microsoft Research India in summer 2012. In collaboration with Research Engineers Andrew Cross and Aditya Vashishtha, and advised by Dr. Bill Thies, Dr. Venkat Padmanabhan and Dr. Ed Cuttrell, we measured traffic density and speed for non-laned heterogenous traffic from videos.
I made a failed startup effort, with Pradeep Banavara, where we tried to design incentivizing Android apps for people to share their GPS traces. Another piece of "outside PhD traffic related work" was done as a research engineer at SMU Livelabs, where under the guidance of Prof. Rajesh Balan, I analyzed GPS traces from a taxi company to detect their operational inefficiencies. A project on vehicle type classification using smartphone sensors, was done by Shilpa Garg, a masters student at IIIT Delhi, as part of her masters thesis under Prof. Pushpendra Singh. I collaborated remotely on this project, and it was a fun experience.
[2] Human mobility measurement, both outdoor and indoor
I measured walkability across world cities using public web APIs in collaboration with Daniele Quercia, from Bell Labs Cambridge. As an intern at MNS group in summer 2011, advised by Dr. Venkat Padmanabhan and Dr. Krishna Chintalapudi, and collaborating with Research Engineer Anshul Rai, we devised a particle filter based method to create indoor Wi-Fi maps automatically from inertial sensor traces obtained from smartphones. At SMU Livelabs, I collaborated with Research Engineer Kasthuri Jayarajah (now a PhD student), under the advice of Prof. Youngki Lee, Prof. Rajesh Balan and Prof. Archan Misra, to detect social groups in real time in urban spaces like shopping malls and airports, from locations and activities inferred from smartphones.
[3] Privacy enhancing systems for smart devices
At MPI, I work on mobile privacy with PhD student Paarijaat Aditya, the computer vision group at MPI-INF (vision collaborators) and guided by Prof. Peter Druschel and Prof. Bobby Bhattacharjee. I am discussing my recent work on hardware security extensions and mobile operating systems with Matthew Lentz, who has prior experience with Android kernel programming and device drivers.
[4] Empirical audit of public policies
The empirical audit of Facebook's Free Basics was tremendous fun in terms of connecting with people. Prof. Dave Choffnes, Prof. Krishna Gummadi, Prof. Ponnurangam Kumaraguru and Prof. Ihsan Ayyub Qazi were some faculty members who collaborated in this project. We needed to run experiments in 15+ developing countries across Asia, Africa, South and Central America. So I worked with numerous students in different countries like Amreesh Phokeer, Taslim Arefin Khan, Hasnain Ali Pirzada, Sohaib Ahmad, Zaid Ahmed Khan, Siddharth Singh, Vedant Nanda and Satadal Sengupta.
Cities in developing countries like India are crumbling under the pressure of population and adhoc infrastructure growth. Delhi air is unbreathable. Bengaluru lakes are spewing toxic froth. Kolkata flyovers are collapsing crushing citizens. I dream my research to play a small role in creating data-driven public debates and dialogues with the authorities, instead of temporary media frenzy after such urban disasters. I wish to join forces with the handful of individuals and companies, working in sectors of green energy, electricity audit, emission reduction, urban waste management etc., to gradually make the Indian cities sustainable and more livable. Vehant Technologies, Zendrive, Zenatix, Altigreen, Nascent and Mapunity are some such Indian startups with whom I collaborate and will continue to do so in future.