Research On FDA-MIMO Radar Target Detection Algorithm In Complex Environments

The Frequency diverse array-multiple input multiple output(FDA-MIMO)radar,as a novel array radar,possesses additional range degrees of freedom(DoFs)compared to phased-array(PA)or MIMO radars.In contrast to traditional FDA radar,it maintains range and angle information while decoupling range and angle together with time invariance.Hence,FDA-MIMO radar holds promising applications in clutter suppression,interference mitigation,target detection and tracking.This dissertation commences its exploration by delving into the analysis of FDA radar array characteristics,meticulously investigating the nature of radar cross section(RCS)and generalized ambiguity function(GAF).Building upon this foundation,the thesis then extends its focus to comprehensive investigations on FDA-MIMO target detection in diverse environments such as clutter,noise,and mainlobe interference.The main research contributions and innovations of this thesis are summarized as follows:(1)This dissertation establishes the RCS model for FDA radar and further defines its GAF function.Specifically,this dissertation commences by establishing an FDA radar RCS model,drawing upon the intricate electromagnetic propagation characteristics.This comprehensive model encompasses both the dumbbell-scatter targets and multi-scatter targets.Within this framework,the profound impact of frequency increments on target RCS is meticulously scrutinized,unveiling insights into the statistical characteristics of FDA radar target RCS.A precise definition of the GAF for FDA radar is also introduced,paving the way for a comprehensive analysis.The exploration then delves into a thorough investigation of GAF’s performance across range,angle,and Doppler dimensions.Furthermore,the intricate interplay of dimensions,such as range-angle,range-Doppler,and angle-Doppler,is meticulously explored.This analysis brings to light a crucial revelation-excessive frequency offset can result in secondary ambiguity within range units,leading to the expansion of the transmitted beam and the emergence of Doppler ambiguity.(2)Addressing the problem of target detection without training data in cluttered environments,this thesis proposes adaptive target detectors that don’t require training data.It also designs robust detectors for situations with space-time-range uncertainties and limited test data,respectively.Due to the range-dependent nature of clutter signals in FDA-MIMO radar,it is challenging to directly utilize received data collecting from adjacent range bins as training data.Hence,this thesis introduces detectors with constant false alarm rate(CFAR)characteristics based on Rao and Wald criteria.We derive closed-form expressions for detection probability(PD)and probability of false alarm(PFA).Considering signal mismatch in radar operation,the dissertation proposes robust subspace detectors based on proj ecting the signal vectors of space-time-range uncertainties onto different subspaces.These new detectors outperform non-subspace detectors in terms of robustness.Given the limited dwell time of FDA-MIMO radar,which results in insufficient test data,this dissertation uses spatial projection(matched filtering)to reduce the dimensionality of the detection problem,introducing three CFAR-based reduced-dimension detectors.These new detectors not only have lower computational complexity but also strong interference resistance,and they have lower data requirements compared to existing detectors.(3)Considering the target detection with training data in noise environments,this dissertation presents multiple adaptive target detectors for both homogeneous environment(HE)and partially HE(PHE).Additionally,this dissertation also includes considerations for robust target detection and distributed target detection,respectively.After undergoing range compensation,the data,including clutter signal collected from adjacent range bins,can indeed be utilized as the training data.Therefore,taking the target detection problem in the presence of training data operating in noise environments(including receiver thermal noise,interference and the clutter after range compensation)into consideration,this dissertation designs three detectors based on the OGLRT,TGLRT,and Rao criteria.We introduce an adaptive matched filter(AMF)detector during statistical analysis for the Rao detector.It derives closed-form expressions for PD and PFA for the proposed detectors and proves that they all possess CFAR characteristics.Addressing target detection problem in PHE scenarios,we design four detectors with CFAR characteristics based on the OGLRT,TGLRT,Rao,and Wald criteria,respectively.Besides,this dissertation proposes a method to enhance detector robustness in cases of signal mismatch,adding a random variable to the effective hypothesis,and designs four detectors.These detectors outperform existing detectors in terms of robustness.The dissertation applies the super-resolution characteristics of FDA-MIMO radar to detect distributed targets within a range cell.Moreover,we establish the received signal expression for distributed targets and design four adaptive detectors with CFAR characteristics.(4)To solve the problem of target detection in the presence of mainlobe deception interference,various adaptive detectors have been firstly proposed based on both non-subspace and subspace methods.This dissertation first establishes a signal detection model for scenarios with mainlobe deception interference and proposes four adaptive detectors based on the OGLRT,TGLRT,Rao,and RD criteria.Numerical simulations validate that MIMO radar cannot detect targets in the presence of main lobe deception interference.Addressing target detection in main lobe deception interference scenarios with a signal mismatch,the dissertation projects the transmit and receive steering vectors onto different subspaces,transforming the original detection problem into a threesubspace detection problem,and designs five adaptive detectors.These detectors exhibit less performance degradation in the presence of signal errors compared to non-subspace detectors.(5)Taking a prior knowledge about the covariance matrix into mind,this dissertation discusses adaptive target detectors based on a Bayesian framework operating in clutter,noise environments,respectively.Specifically,we incorporate prior knowledge of clutter into detector design,resulting in a Bayesian detector that doesn’t require training data.These detectors exhibit significantly improved performance at low SNR and enhanced robustness.Furthermore,in noise environments,this dissertation jointly considers training data and prior knowledge of the covariance matrix in detector design.The proposed detectors enhance resistance to signal mismatch and no longer depend on training data.For the problem distributed target detection in cluttered environments with prior knowledge,the dissertation assumes that the covariance matrix follows an inverse Wishart distribution,and designs Bayesian detectors based on the GLRT,Rao,and Wald criteria.