In this paper, we study optimal radar deployment for intrusion detection,with focus on network coverage. In contrast to the disk-based sensing model ina traditional sensor network, the detection range of a bistatic radar dependson the locations of both the radar transmitter and radar receiver, and ischaracterized by Cassini ovals. Furthermore, in a network with multiple radartransmitters and receivers, since any pair of transmitter and receiver canpotentially form a bistatic radar, the detection ranges of different bistaticradars are coupled and the corresponding network coverage is intimately relatedto the locations of all transmitters and receivers, making the optimaldeployment design highly non-trivial. Clearly, the detectability of an intruderdepends on the highest SNR received by all possible bistatic radars. We focuson the worst-case intrusion detectability, i.e., the minimum possibledetectability along all possible intrusion paths. Although it is plausible todeploy radars on a shortest line segment across the field, it is not alwaysoptimal in general, which we illustrate via counter-examples. We then present asufficient condition on the field geometry for the optimality of shortest linedeployment to hold. Further, we quantify the local structure of detectabilitycorresponding to a given deployment order and spacings of radar transmittersand receivers, building on which we characterize the optimal deployment tomaximize the worst-case intrusion detectability. Our results show that theoptimal deployment locations exhibit a balanced structure. We also develop apolynomial-time approximation algorithm for characterizing the worse-caseintrusion path for any given locations of radars under random deployment.
展开▼
