Decentralised detection strategies under communication constraints

Decentralised detection systems with parallel decision fusion topology have received significant attention, especially for surveillance radar systems. Typically, a deterministic test (DT) is employed, and the optimal local decision strategies and the fusion rule are determined according to the Neyman-Pearson (N-P) criterion. The authors show how, in practical applications, the communication bottle-neck may significantly limit the detection performance of distributed systems employing a conventional DT. A paradigm is introduced for decentralised detection under communication constraints, based on the concept of dependent randomisation (or scheduling) of the local decision and fusion rules. Numerical results show that the proposed approach provides improved performance over decentralised detection strategies based on deterministic tests, with a moderate increase in system complexity; the improvement is particularly noticeable in spiky (non-Gaussian) clutter environments