Study On Parameter Estimation And Data Fusion Of MIMO Radar

Multiple-input multiple-output(MIMO)radar exploits waveform diversity,spatial diversity,and polarization diversity technology.Compared to the traditional radar,MI-MO radar has more degrees of freedom and more advantages in parameter estimation,target detection and recognition.According to the difference in antenna place,MI-MO radar can be classified into collocated MIMO radar and distributed MIMO radar.Based on the MIMO radar system,this dissertation addresses the fast parameter esti-mation with collocated MIMO radar and data fusion with distributed MIMO radar.The main contributions are as follow.1.To reduce the computational complexity in parameter estimation with MIMO radar,a fast joint direction of departure(DOD)and direction of arrival(DOA)estima-tion method is proposed.Based on a bistatic MIMO radar geometry with one transmit array and two subarrays at the receive array,the cross-correlation matrix(CCM)is con-structed.Combined the block matrix and rotational invariance technique,the direction finding is accomplished.This method does not need to perform eigendecomposition,the estimated DODs and DOAs can be paired automatically,thereby dramatically re-duce the computational burden.2.Traditional uniform circular array ESPRIT(UCA-ESPRIT)can deal with the coupling between the elevation angle and azimuth.However,this method has two limitations.In the case of the radius of UCA equals to transmitting wave length,the elements of UCA must larger than 12,and the maximum recognition targets are 5.To deal with the limitations,a low computation complexity algorithm for single source direction finding of MIMO radar with uniform circular transmit array is presented,which has no restriction in the number of elements for the UCA.Based on the rota-tional invariance property of receive array,the DOA is estimated.Then the estimated DOA is substituted into the steering vector of receive data,and the steering vector of transmit UCA is extracted.In the case of single source,the phase of covariance ma-trix is extracted and transformed to matrix form,based on the least square method,the close form of DOD is obtained.On the other hand,a new direction finding method for MIMO radar with non-circular sources is proposed.The method takes advantage of the properties of non-circular sources to formulate a new virtual array.The direction finding with MIMO radar is transformed to the direction finding with UCA,the non-circular UCA-ESPRIT(NC-UCA-ESPRIT)is applied to estimation the elevation and azimuth of target.3.In modern data fusion system,the number of sensors is always fixed,and the signal is collected uninterruptedly.To deal with this practical case,an efficient multi-sensor data fusion algorithm using random matrix theory(RMT)is proposed under the additive Gaussian noise model.The fusion result of the proposed algorithm can be calculated in one iteration using the RMT.The proposed algorithm is an approximate implementation of the linear minimum mean squared error(LMMSE)estimator that is not directly realizable.When the dimensionality of signal is larger than the number of sensors,the proposed estimator,as a biased estimator,has a lower mean squared error(MSE)than the maximum likelihood estimation(MLE)fusion algorithm by trading off the bias and the variance.Theoretical calculation and simulations verify the per-formance of the proposed fusion algorithm is close to the normalized LMMSE for any linear estimator.4.Traditional detection method using constant false alarm probability can not de-tect the faint target,a data fusion method using the Markov chain is proposed for dis-tributed MIMO radar.Based on the statistic characteristics of echo data,and combined the motion features of target,the state transition probability matrix is computed.Com-pared with the threshold,the measuring point is judged,and the real target is detected.The simulation and outdoor experiment results show that this algorithm can distinct-ly reduce detection threshold without increasing the false alarm probability.Thus the system detection performance can be effectively improved.5.The above Markov chain based data fusion method need to compute the first and last locations of target,therefore is unstable and the calculating of detection thresh-old is time consuming.Based on the Baysian estimation,a data fusion method for distributed MIMO radar is proposed.The prior probability is obtain from consecu-tive frames,combined the restricted maneuverability of moving target,the posterior probability is computed based on the Baysian theory.The related path is built among measuring points,the probability of moving target is then estimated.Finally,the re-lated pairing is accomplished by using threshold judgment.The proposed method can simplify the calculation of detection threshold,which is robust.The outdoor experi-ment with aircraft data is completed,the results show that the moving target can be judged clearly and farther.