The success of emerging sensor network technologies rests on the ability to detect events of interest. In a setting where local sensors preprocess observations before transmitting data to a fusion center, the corresponding decision making problem is termed decentralized detection. Decentralized detection has found applications in sensor networks. Classical decentralized detection framework has limited application to modern wireless sensor networks. Reevaluating classical decentralization is an instrumental step in deriving valuable guidelines for the efficient design of sensor networks. Departure from the classical decentralized detection framework comes from the realization that wireless sensors transmit data over a common wireless spectrum. However, sensor nodes are bounded by design factors like cost, spectral bandwidth and and power requirements. A natural initial approach to the capacity-constrained problem is to overlook the specifics of these physical parameters and to constrain the sum-capacity of the multiple-access channel available to the sensors. Recent initiatives on decentralized detection consist in incorporating the effects of the wireless environment on the transmission of messages between the sensors and the fusion center. Also, researchers have started to explore new paradigms for detection over wireless sensor networks, which are based on the view that future systems are application-specific systems. Another paradigm may use local message passing, which reduce the need for spectral bandwidth and there is no designated fusion center. Another way is for the nodes to exploit a multihop communication scheme where data packets are relayed from sensor to sensor until they reach their respective destinations.
ASJC Scopus subject areas
- Signal Processing
- Electrical and Electronic Engineering
- Applied Mathematics