Research On Adaptive Detection Of MIMO Radar With Widely Separated Antennas

Multiple-input multiple-output (MIMO) radar with widely separated antennas hascaught much attention in recent years. It refers to an architecture that employsmultiple transmitters and receivers, capturing the spatial diversity of the target’sradar cross-section (RCS) to improve system performance. The adaptive detectionof MIMO radar against the compound-Gaussian clutter and the optimization designbased on detection performance are discussed in this paper.Firstly, the signal model of MIMO radar is established. Two kinds of constantfalse alarm rate (CFAR) detectors are proposed. The analytic expressions of theirfalse alarm probabilities are derived to prove their CFAR property in theory,respectively. The good detection performance of both detectors is verified by thesimulations.Secondly, the measurement and statistical models within the generalizedmultivariate analysis of variance (GMANOVA) framework are introduced incompound-Gaussian clutter with inverse gamma texture. The PX-EM algorithm isapplied to estimate the unknown signal and clutter parameters. A statistical decisiontest using these estimates is developed. Based on the statistical characteristics ofthis test, an algorithm to adaptively distribute the transmitted energy among thetransmitters and maximize the detection performance is proposed. Simulationanalysis demonstrates advantages of the MIMO setup and adaptive energyallocation in target detection.Finally, introducing polarization diversity, the signal model of polarimetricMIMO radar is set up against the compound-Gaussian clutter. A polarimetricadaptive detector (MAP-GLRT) based on maximum a posteriori estimation and thetwo-step generalized likelihood ratio test is presented. Its CFAR property is verifiedtheoretically. Based on the basic theories of game theory, the polarizationoptimization design of MIMO radar is formulated in the form of a two player zero-sum game between an opponent and the radar design engineer. The optimal purestrategy and the optimal mixed strategy are solved. Compared with theconventional approaches, this approach does not require estimation from the training data, and hence it is less complex and easy to implement in a practicalsystem. The influencing factors, the performance advantages of MAP-GLRTdetector and the effectiveness of the optimization algorithm are analyzed by thesimulation results.