Collector Receiver Design for Data Collection and Localization in Sensor-driven Networks

We consider a sensor network in which the sensors communicate at will when they have something to report, without prior coordination with other sensors or with data collection nodes. The burden of demodulating the sensor data, and localizing the sensor which is communicating, falls on a network of collector nodes which are perpetually monitoring transmissions from the sensor network. This model allows the random deployment of very large numbers of sensor nodes with minimal capabilities, while shifting the complexity to a network of collector nodes. While the philosophy is similar to prior work on "imaging" sensor nets, the key difference is that the communication model is now sensor-driven, rather than collector-driven. The two major technical challenges addressed in this paper are as follows: (a) Are there simple physical layer implementations of the collector receiver for jointly solving the tasks of detection of a sensor transmission, estimation of the direction from which it comes, and demodulating the data? (b) Given that the collectors are not time synchronized well enough to permit the use of time-difference-of-arrival techniques for sensor localization, how well can the sensors be localized with spatial information alone, assuming that each collector node has a relatively small number of antennas? The results reported in this paper indicate that the preceding issues can be addressed satisfactorily with appropriate design of the collector physical layer, together with Bayesian combining of the spatial information extracted by each collector.