A sensor network is a spatio-temporal sampling device with a wireless communication infrastructure. This poses a number of interesting questions on sampling, representation, compression and communication as well as on applications.
In this talk, we start with a short overview of the Center on Mobile Information and Communication Systems (http://www.mics.org), which is a Swiss national center of competence on ad hoc networks, sensor networks, and applications in environmental monitoring. We then move to a series of challenges and a set of results.
First, we consider the spatio-temporal structure of distributed signals, with an emphasis on the physics behind the signals, and results on sampling. In particular, we develop the notion of the plenacoustic function, inspired by the plenoptic function, and give sharp sampling results on acoustic fields.
Second, we address distributed source compression, in particular in the lossy case. Using the structure of distributed signals, we derive distributed rate-distortion results, showing the rate-distortion function for acoustic fields.
Third, we consider the interaction of distributed source compression and transmission, with a particular focus on joint source-channel coding. We review a simple yet emblematic case where a simple analog, uncoded system is exponentially better than a classic, separation based method. Then, we move to applications in environmental monitoring, an interesting challenge for wireless sensor networks. We describe a tomographic sensing method for temperature and wind measurement which has a positive scaling law.
Finally, we describe SensorScope (http://sensorscope.epfl.ch), an experimental campus wide weather station network co-developped with environmental engineers for urban climate measurement and modeling.
This is joint work with T.Ajdler, G.Barrenetxea, H.Dubois-Ferriere, R.Konsbruck, I.Jovanovic, E.Teletar and M.Parlange (EPFL), P.L.Dragotti (Imperial) and M.Gastpar (UC Berkeley). The work is sponsored by the Center on Mobile Information and Communication Systems (www.mics.org), funded by the Swiss National Science Foundation.