Research On AOA Estimation Of Heterogeneous Polarization Sensitive Array Enhanced By Nonlinear Observation

The spatial signal angle of arrival(AOA)estimation problem is a fundamental problem in array signal processing.In recent years,the antenna array(AA)performance requirements for direction-finding systems have been getting higher and higher and the size of AA is getting more prominent to improve spatial resolution.However,problems such as expensive hardware cost,high system complexity and high power consumption have emerged.Low-power nonlinear radio frequency(RF)links can be introduced into conventional AA to reduce power consumption and hardware cost.However,at the same time,traditional AOA estimation algorithms based on complex observations will fail when these nonlinear components introduce nonlinear observations into the direction-finding systems.Developing an AOA estimation algorithm that can cope with nonlinear observations is necessary.In addition,there is often an error in the direction-finding system,making the array manifold deviate to a certain extent.Then the performance of most AOA estimation algorithms will seriously deteriorate or even fail.This topic introduces the low-power and low-cost magnitude RF chain into the traditional PSA.An efficient and energy-saving magnitude-aided PSA(MA-PSA)is designed.The problems of incoherent AOA estimation,coherent AOA estimation and magnitudephase error correction in the MA-PSA system are thoroughly investigated.The main results of the work are as follows.1.An iterative maximum likelihood(IML)AOA and channel response joint estimator for MA-PSA is proposed.In IML,the MUSIC obtains an initial AOA estimate based on the complex observations provided by the conventional PSA array.Then,in each iteration of IML,each angle value of the AOA estimate is updated by the single angle search(SAS)method.In SAS,the neighborhood of the target angle is uniformly gridded to obtain several pending angles.Since the posterior distributions of the unknown parameters concerning the complex and magnitude observations are the complex Gaussian and Rayleigh distributions,the expressions of these two distributions can be used to calculate the likelihood function values and update the target angle values to this ML estimate.IML executes the SAS method alternately until the AOA estimate changes slowly enough between two adjacent IML iterations.When the IML algorithm iteration terminates,the final AOA and channel response estimates are output.The simulation results show that the magnitude observations introduced in MA-PSA can improve the AOA estimation performance of the IML and the IML does not need to use a reference source.In addition,introducing more magnitude observations can improve the performance of the IML.The use of SAS leads to a significant reduction in the complexity of the IML relative to the conventional ML.2.A hybrid Gibbs joint estimator for coherent source AOA and channel response in MA-PSA is proposed.The hybrid Gibbs first obtains the initial estimate of the coherent source AOA using the signal subspace fitting(SSF)algorithm based on the complex observations provided by the PSA and calculates the initial estimate of the channel response using the least squares(LS)method.The initial estimates of AOA and channel response are used as the initial samples of the hybrid Gibbs estimator.Due to the introduction of magnitude observations,the full conditional distributions of the AOA and channel response parameters are not standard and too complex to be directly sampled during the Gibbs sampling process.Therefore,samples are drawn from the predetermined proposed distributions in each iteration.Their acceptance probabilities are calculated to determine whether to retain the samples,following the MetropolisHastings Markov chain Monte Carlo(MH-MCMC)method.The invariant distribution of the Markov chain formed by these samples is the full conditional distribution of the unknown variables about the observation,so the mean value of the samples after the burn-in period can be used as the final mixed Gibbs estimate of the AOA and the channel response.Simulation experiments show that the magnitude observation introduced in MA-PSA can improve the AOA estimation performance of the hybrid Gibbs and the hybrid Gibbs does not require a reference source.The performance of hybrid Gibbs is verified in the indoor multipath channel model of WINNER-II.3.A hybrid Gibbs AOA and magnitude-phase error joint estimator for MA-PSA is proposed.The algorithm first uses an auxiliary calibration source with known AOA to compute an initial estimate of the magnitudephase error using the LS method.The initial value of AOA is estimated using the calibrated MUSIC algorithm.The initial AOA and magnitudephase error estimates are used as the initial samples for the hybrid Gibbs estimator.Due to the introduction of magnitude observations and array magnitude-phase errors,the full-conditional distribution of AOA and magnitude-phase errors about the hybrid observations must be too complex to be sampled directly by Gibbs.So now,the proposed distribution of MHMCMC can still be used to draw samples.Therefore,the mean value of the sample after the burning period can be used as the coherent source AOA and magnitude-phase error estimate of the hybrid Gibbs.The simulation experiments verify the effectiveness of the hybrid Gibbs algorithm for j oint AOA and magnitude-phase error estimation from several perspectives.The simulation experiments show that the hybrid Gibbs estimator based on hybrid observations has significant gains over the MUSIC and LS that obtain initial AOA and magnitude-phase error estimates.The hybrid Gibbs algorithm also outperforms the MH-MCMC algorithm.