Research On Target Parameter Estimations And System Design For MIMO Radar Array

This thesis has investigated two theoretical problems involved in Multiple-InputMultiple-Output (MIMO) radar systems. One is why MIMO radar should be devel-oped; the other is how to design a MIMO radar system. First of all, two fundamentalissuesareconcerned,i.e.,correlationcharacteristicofchannelsandambiguitycharacteris-tic. The relationship between the correlation characteristic of channels with radar systemparameters is obtained, and the conventional ambiguity function is generalized; secondly,two new maximum likelihood (ML) direction finding (DF) algorithms using expectation-maximization (EM) and space alternating generalized expectation-maximization (SAGE)are proposed; thirdly, a framework of theory on MIMO radar target parameter estima-tion performance analysis is constructed, which not only includes the fundamental per-formance guidelines, such as the direction finding accuracy, thresholds of detection andresolution, but also includes fundamental performance measurements, such as the systemidentifiability capacity, manifold ambiguity and sensitivity analysis; finally, according tothe aforementioned performance framework, the corresponding system design strategiesare introduced.Because of the application of waveform diversity technique, the obtained waveformdiversity gain for MIMO radar with colocated antennas is so excellent that the huge po-tentialonparameterestimationandparameteridentifiabilitycanbestemmedfrom. Basedon the comprehensive and systemic review on the state of art of MIMO radar, the workof this thesis includes three parts as follows.Thefirstpart(ELEMENTS)mainlyconcernsMIMOradarchannel, ambiguityfunc-tion and ML DF algorithms. The general signal model of MIMO radar is introduced first,which is the foundation of posterior work. Borrowing the notion “channel” in commu-nication theory, the relationship between correlation characteristic of channels in MIMOradar and system parameters with measurement model is focused on, which demonstratesthat the higher carrier frequency, larger target, shorter detecting distances and larger an-tenna spacing will get the lower correlation characteristics of channels. Note that thetraditional ambiguity function proposed by Woodward considers the ambiguity propertyjust from the transmitted signals, which is obviously hard to describe the practical reso-lution capability. Then, two generalized ambiguity function based on matching filtering and maximum likelihood function are proposed from the received signals at the receivers,furthermore, theeffectsofbistatictrianglegeometryontheambiguitypropertyaremainlyanalyzed. Ultimately, two ML DF algorithms using EM and SAGE are proposed, besides,their convergence analyses are given, and the conclusion that the asymptotic convergencerate of SAGE algorithm is prior to one of EM algorithm is proved.Inthesecondpart(THEORY),thetheoryperformanceanalysisframeworkofMIMOradar target parameter estimation is constructed based on the theories of differential ge-ometry and system identification. First of all, MIMO radar virtual array is mapped intothe corresponding manifold space via steering vector. Then the DF accuracy, thresholdsof detection and resolution, and manifold ambiguity conclusions of conventional arraysignal processing are generalized to MIMO radar systems. In the viewpoint of systemidentification, taking dimension theory as a tool, the parameter identifiability capabilityofcolocatedMIMOradar is concernedinthenarrowbandscene, then, theparameteriden-tifiability capabilities both of colocated and separated MIMO radars are investigated inwideband scene, according to the fundamental flow of wideband signal processing. Fi-nally, the DF sensitivity analyses relative to the antenna position uncertainties (APU) areconcernedfromtwo pointsof view. Oneis the measurements ofsensitivityand rank-1DFambiguity based on maximum likelihood algorithm. The other is the antenna importantfunction (AIF) and overall system sensitivity (OSS) derived from MIMO radar virtualarray manifold surface, regardless of any certain DF algorithm. The theory frameworkof MIMO radar target parameter estimation performance analysis can be as the guidelineof system design, also can be as the tool to discuss the DF property for given systems,at the same time, can be as the criterions to evaluate the employed algorithms for theirindependence of certain algorithms.According to part two, the property of MIMO radar target parameter estimationmainly depends on the its antenna geometry, moreover, it is obvious that the transmit-ted signals are very important for system capability. The third part (APPLICATIONS)focuses on the issue of MIMO radar system design, including the strategies both of an-tenna geometry configuration and transmitted signal set. Considering that the coupledparameter estimation degrades the performance gravely and that the uncoupled directionestimation is equivalent to all elements of Fisher information matrix (FIM) except the el-ements in diagonal being zero, the sufficient and necessary antenna position conditions of uncoupleddirectionestimationarerespectivelyderivedforbistaticandmonostaticMIMOradar. Furthermore, since the uncoupled direction estimation condition can not guaranteethe optimum antenna geometry in the sense of Cramer-Rao Bound (CRB), the new cri-terion is proposed, which is to maximize the parameter estimation performance with thegiven number of transmit and receive antenna and a certain restricting region. Two cases,usually encountered in practical applications, are considered. With no prior direction in-formation, the isotropic antenna geometry is the optimal one, then, a quantitative resultswhich the antenna positions should be fulfilled is obtained. Alternatively, for a directiveantennageometrywithapriordirection,theTrace-Optimal(TO)criterion(minimizingthetrace) on the average CRB matrix can be used and a qualitative explanation for antennageometry strategy is proposed using simulated annealing (SA) algorithm. Finally, in theviewpoint of information theory, radar systems accomplish their mission based on theinformation capability. Then, addressing the problem of MIMO radar system design fortargetidentificationandclassification, themaximumchannelcapacityisusedascriterion,which includes the strategies both of antenna geometry and transmitted signals.In conclusion, this thesis, mainly focusing on the MIMO radar DF, have constructedthe theory framework of target parameter estimation property analysis, including DF ac-curacy, thresholds of detection and resolution, manifold ambiguity, parameter identifia-bility capacity and sensitivity analysis, and have proposed several system design strate-gies, including uncoupled direction estimation, the sense of optimal CRB and maximumchannel capacity. This thesis has responded across-the-board to two theoretical problemsbrought farward at the beginning. It is worth to point out that the methodology presentedinthisthesiscanbegeneralizedtotherelativefundamentalinvestigationbothofseparatedMIMO radar and array signal processing.