Information-Theoretic Treatment of Sensor Data Collection: A Perspective on Processing and Communication Tradeoffs

In sensor networks, a device may operate on fuzzy inputs, thereby generating a result that possibly deviates from the reference datum in physical world being sensed. The extent of deviation (i.e., inaccuracy of data) and the time it takes to compute the output result (i.e., timeliness of data) depend on the number of orthogonal information elements processed from the sensed data. Due to the large dimensionality of input data, a sensor device may need to incur significant amounts of computation on a data and communication among with other devices to exchange pre-processed data. Since the sensor devices are often resource-constrained (i.e., have limited amount of processing cycles, memory sizes, and battery power), an important issue is the amount of computation and communication cycles a device can spend during data processing. The paper provides an information-theoretic discussion of the tradeoffs between the resources expended by a device to process its input data and the timeliness and accuracy of its output result. The goal is to provide quantitative methods of evaluating the tradeoffs, so that sensor devices can adapt their operations based on the resource availability.