The study of networked control systems requires a convergence of tools from information theory, coding theory, computer science and control theory. Emerging applications of networked control systems include autonomous vehicles, intelligent highway systems, networked city services, sensor networks and smart grid. I am interested in the signal processing, coding and control theoretic challenges involved in the design and analysis of such systems. More details can be found in my Research Statement.
Error Correcting Codes for Interactive Communication
Traditional information theoretic notions like block error probability and Shannon capacity do not provide the right means to measure the reliability of communication channels that are in the feedback loop of control systems. The appropriate notion of reliability is what is known as anytime reliability and conventional techniques of block and convolutional coding do not achieve this reliability. The kind of codes that achieve anytime reliability are the so called tree codes. Tree codes are also central to the problem of simulating interactive protocols over noisy channels. Even though the notions of tree codes and anytime reliability have been around in the works of Schulman and Sahai for almost two decades now, there have been no practical constructions of tree codes till date. For the first time, we have an explicit ensemble of tree codes that are efficiently decodable over erasure channels. Erasure channels are used to model errors in most of the existing communication links
"Error Correcting Codes for Distributed Control", working paper. arXiv
"Anytime Reliable Codes for Stabilizing Plants over Erasure Channels", IEEE International Conference on Decision and Control (CDC), Orlando, Florida, Dec 2011. Paper
"Linear Error Correcting Codes with Anytime Reliability", IEEE International Symposium on Information Theory (ISIT), St. Petersburg, Russian, Aug 2011. Paper
Data Fusion over Sensor Networks
Recent advances in VLSI and MEMS technology have led to the availability of cheap, low quality and low power consumption sensors in the market. This has generated a great deal of interest in wireless sensor networks due to their potential applications in several diverse fields. One of the important challenges in deploying sensor networks is the fusion of quantized data collected from the sensors at the fusion center. The corresponding estimation problem is non-linear and traditional approaches like the kalman filter are not applicable. In this context, we developed an efficient particle filtering technique that exploits the underlying state space structure of the ambient process. The resulting technique which we call the 'Kalman Like Particle filter' delivers close to optimal performance using an order of magnitude fewer particles than other contemporary approaches that apply a particle filter directly to the original problem. The resulting work that was submitted to CDC in 2009 was one of the four finalists for the best student paper award.
"The Kalman Like Particle Filter : Optimal Estimation With Quantized Innovations/Measurements", submitted to IEEE Transactions on Signal Processing. Paper
"The Kalman Like Particle Filter : Optimal Estimation With Quantized Innovations/Measurements", IEEE International Conference on Decision and Control (CDC), Shanghai, China, Dec 2009. Paper
"Particle Filtering for Quantized Innovations", International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Taipei, Taiwan, April 2009. Paper