Study On Angle Estimation In Bistatic MIMO Radar Based On Real-valued Processing

Compared with the traditional phased-array radar,multiple-input multiple-output(MIMO)radar is a new radar system and has a lot of advantages,which provides new development opportunity for radar system.Bistatic MIMO radar combines the advantages of both MIMO radar and bistatic radar,which attracts more and more attention.Combining the tensor decomposition technique with the spatial filtering technique,this dissertation puts main focus on the direction-of-departure(DOD)and direction-of-arrival(DOA)estimation problem in bistatic MIMO radar.The main contributions of this dissertation are shown as follows:(1)The traditional angle estimation methods for bistatic MIMO radar are based on the signal models in the form of matrix,which do not exploit the multidimensional structure of received data.Through real-valued processing,a tensor-based unitary parallel factor(U-PARAFAC)method is proposed.Firstly,we prove that the real-valued tensor signal model after real-valued transformation follows a PARAFAC model.Once the uniqueness condition is satisfied,the real-valued loading factors that include angle information can be obtained subsequently.Since the proposed method avoids beamspace estimation,the disadvantages of the traditional subspace-based methods can be overcome.Simulation results verify that the proposed method has better estimation accuracy especially in dealing with closely spaced and more than two correlated targets.(2)The traditional bistatic MIMO radar employs full waveform diversity and has flat transmit coherent gain in all directions.However,the general spatial sector where the targets are located is usually known especially for target tracking.Under this condition,it will cause undesirable energy waste when still using the orthogonal transmit waveforms.To solve this problem,a novel angle estimation method for transmit beamspace(TB)based bistatic MIMO radar is proposed,which utilizes matrix signal model.Firstly,the TB matrix is designed to focus the transmitted energy within the desired spatial sector.Furthermore,the transmitted energy distributed in the out-of-sector should also be minimized.The CRB(Cramer-Rao bound,CRB)of TB-based bistatic MIMO radar for DOD and DOA estimation is derived,which is lower compared with the traditional bistatic MIMO radar.(3)In TB-based bistatic MIMO radar,the transmitted energy focusing technique is combined with PARAFAC tensor decomposition technique.Firstly,the received data structure of the TB-based bistatic MIMO radar is different from the traditional bistatic MIMO radar.In order to transform the complex tensor model into real-valued tensor model,the matrix structure of the TB matrix must be properly designed.For this purpose,the equality constraint is added in the traditional optimization model.Once the structure condition of TB matrix is satisfied,the complex tensor model can be transformed into real-valued tensor model through the traditional ways.Under this precondition,the obtained real-valued tensor model is proved to follow a PARAFAC model.(4)In the traditional bistatic MIMO radar,spatial filtering is performed in both transmit and receive side and a joint real-valued beamspace-based method for angle estimation is proposed in this dissertation.The mainlobe-to-sidelobe ratio(MSR)of the traditional Discrete Fourier Transform(DFT)spatial filter is low.Furthermore,once the beamspace dimension is identified,the bandwidth and sidelobe level of the DFT beamspace filter cannot be flexibly controlled as required,which will greatly decrease the MSR.Different from the traditional DFT beamspace filter,the proposed filter is designed through convex optimization,which greatly improves the MSR.Most importantly,the structure of TB and receive beamspace(RB)matrix can also be properly designed through convex optimization,which is indispensable in constructing the real-valued signal model.(5)In the traditional bistatic MIMO radar,the above mentioned spatial filtering technique is respectively combined with Higher-order SVD(HOSVD)and PARAFAC techniques and the corresponding angle estimation methods are respectively proposed.After constructing the real-valued tensor model,the signal subspace by performing HOSVD is more accurate than the matrix-based methods.Furthermore,the real-valued tensor model after spatial filtering is still a PARAFAC model and can be decomposed through PARAFAC method.