Distributed GLRT-based Detection of Target in SIRP Clutter and Noise

Abstract

A generalized likelihood ratio test (GLRT)-based distributed detection of a target is proposed for a radar system consisting of multiple radar receivers in the presence of clutter, which is modeled as a spherically invariant random process (SIRP). Based on its local measurements, each radar node solves optimization problems to maximize likelihood functions under null and alternate hypotheses to obtain the GLRT function value, considering that the covariance matrix of the Gaussian process representing the clutter is known, and the texture and receiver noise variance are unknown. The neighboring radars share their local GLRT values and update them iteratively until a convergence is reached to achieve a consensus on the system wide GLRT function value. In the case of noiseless communications links, it is shown that the consensus algorithm achieves global GLRT value, which is used to demonstrate probability of detection versus signal-to-clutter and noise ratio for a given probability of false alarm (PFA). For the noisy links, it is shown that radars may not achieve a consensus and the smallest GLRT value among nodes is considered as the global value. The effect of noise variance on the probability of detection for a given PFA is illustrated with simulation results.