Detection, Mode Selection, and Parameter Estimation in Distributed Radar Networks: Algorithms and Implementation Challenges

Abstract

Target detection, parameter estimation, and tracking capabilities of a radar system can be significantly enhanced by deploying a distributed network. In this article, we consider a network without a central coordinator, wherein only neighboring nodes share limited target related information. We demonstrate that target detection and estimation of range, Doppler, and angle parameters can be significantly improved with the distributed network compared to the localized single-node system. The nodes that observe low signal-to-noise ratio (SNR) values can improve their detection and estimation capabilities considerably by cooperating with others that observe high SNR values. However, when nodes observing low SNR begin to have adverse effects on the global detection performance, these nodes should then be confined to passive sensing, yielding resources to the better performing nodes. We propose a deep learning approach with long-short term memory (LSTM) networks so that each node can compensate neighboring nodes’ observations with respect to its own. Several methodologies for addressing challenges, such as joint data association and parameter estimation for efficient multitarget tracking, distributed detection, and estimation in cluttered radar environments, optimizing network resource allocations, increasing algorithm robustness against channel impairments, and implementing synchronization approaches, are discussed for the distributed radar networks.