Decentralized Underwater Target Detection and Localization
Hu Li, Xiaodong Wang, Shilian Wang
Abstract
This article considers the detection and localization of an underwater target by a sensor network consisting of an active acoustic source, multiple distributed passive sensors, and a fusion center (FC). As the reflected signal from the target could arrive with various directions, a set of collaborative sensors increase the detection probability. We first introduce the centralized generalized likelihood ratio test (GLRT) detector that utilizes the perfect information from sensor observations. In the case of a successful detection, this detector also provides the maximum likelihood estimate (MLE) of the target location. In order to meet the stringent power and bandwidth constraints in underwater sensor networks, we then propose the decentralized GLRT scheme, where sensor observations are quantized into a few bits of data and then transmitted to the FC. The quantization thresholds are optimized by maximizing the Kullback-Leibler divergence between the null and alternative hypotheses. While the existing scan statistic based detector assumes that the sensor array which can observe the reflected signal, is contained in a rectangle area, our proposed decentralized GLRT scheme incorporates the signal reflection model into the detector, and estimates the area of affected sensor array, thus achieving higher accuracy. Finally, we extend the decentralized GLRT detector to the scenario where data are transmitted over error-prone channels, which is not considered in the existing studies. Numerical results demonstrate that the decentralized GLRT detector with only a small number of quantization bits can achieve a similar detection performance as the centralized GLRT.