Sensor networks are often used to monitor phenomena of interest such as the temperature fields, for tracking and detection. In a network of spatially distributed sensors, however, there might not be a central node or base station that can collect all the data acquired by the sensors and perform a centralized estimation. Instead, leader nodes may collect only local information and have to perform local processing in order to infer the distribution of the field of interest or to perform some form of estimation and inference.
The fused data may then be sent to another leader node that will perform a new fusion.
Distributed signal processing algorithms are needed when sensors are distributed and communication is critical, but are also necessary to improve scalability of the sensor network and resilience of the network to sensors failures. Open issues to be addressed include: understanding the best trade-off between local processing versus transmission of raw data to a central node; the best strategy for exchange of local information; convergence rate of local processing algorithms.