Performance and fault tolerance of distributed detection networks

Thesis/Dissertation · Thu Jan 01 04:00:00 EST 1987

OSTI ID:5145924

Reibman, A R

Design, performance, and fault tolerance of distributed detection networks are presented. The optimal design of each local processor in the network is shown to be the likelihood ratio. The local thresholds are computed from coupled equations related by functions of the local processor false alarm and detection probabilities. Network performance is also expressed using local processor performance. Hence, the optimal network performance may be computed knowing only the local performance. A series of studies on the performance of distributed detection networks are made. The first study examines the performance improvement available when the optimal design for a given network is used. Intuitive choices of the optimal fusion design are shown to yield suboptimal performance. The remainder of the studies examine the tradeoff of network structure on the performance of optimally designed systems. The performance of serial and fusion networks are compared both for small and large networks. The performance of the networks is also examined when processor faults may occur. The optimal fault-tolerant design is obtained, both for the distributed networks and the centralized system. The performance improvement of the fault-tolerant system is presented.

• Research Organization: Duke Univ., Durham, NC (USA)
• OSTI ID: 5145924
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200* -- Mathematics & Computers
ARRAY PROCESSORS
COMPUTER NETWORKS
COMPUTERS
DESIGN
DETECTION
DIGITAL COMPUTERS
FAULT TOLERANT COMPUTERS
OPTIMIZATION
PERFORMANCE