Research On DOA Estimation For Multiple-carrier MIMO Radar With Sparse Array

Direction of arrival(DOA)estimation is an important application of array signal processing,the two key characteristics of DOA estimation are the angular resolution and degree of freedoms(DOF)of the array.Usually,the angular resolution strongly depends on the aperture size of the array while the DOF is highly restricted to the number of the sensors in the array.In order to improve the angular resolution under the condition of Nyquist sampling theorem,a simple solution to improve the angular resolution and DOF is to use more physical sensors in the array.However,it improves the cost of the system.To obtain a satisfying angular resolution with a limited number of physical array elements,on one hand,a sparse nonuniform linear array(SNLA)is always preferred,which provides a larger array aperture as well as more DOFs.On the other hand,multiple-in and multiple-out(MIMO)configuration can be used to provide an equivalent array with an extended array aperture and an increased number of virtual sensors.Furthermore,a sparse uniform linear array(SULA)using multiple carrier frequencies can also be utilized to improve the angular resolution.These three different array configurations can be taken as three special cases of a multiple frequencies MIMO radar.In this paper,the DOA estimation for multiple frequencies MIMO is studied and the main results of the dissertation are summarised as follows:1)The equivalent array of the multiple frequencies MIMO radar is studied and the mechanism of improving the spatial angular resolution and degree of freedom of radar system is also analyzed.Based on characteristics of the equivalent array of MIMO for each working frequency,the characteristics of the equivalent array of the multiple frequencies MIMO is further analyzed.To deal with the DOA estimation problem for equivalent sparse linear arrays(SLAs),a DOA algorithm based on root-MUSIC is proposed.By taking the SLA as a filled uniform linear array(ULA)where some elements are omitted,the modified root-MUSIC algorithm is developed for SLA based on the relationshipe between the SLA and its corresponding ULA,which successfully extends the coventional root-MUSIC to SLAs.2)Based on the characteristics of the equivalent array of the SLA in the coprime domain,a compressing sensing based array interpolation algorithm is proposed to performe the DOA estimation with a single snapshot.Firstely,the compressing sensing(CS)technique is used to obtain the coarse initial DOA estimation,then a modified iterative initial DOA estimation based interpo lation algorithm is utilized to obtain the final DOA estimation.The proposed method can efficiently improve the DOA estimation performance.Compared with the CS method,it has no mismatch problem and can deal with more targets than the conventional array interpolation algorithm.The proposed method can make most of the array aperture and the DOFs provided by the virtual array of the SLA in the coprime domain.Furthermore,we also provide two extended sparse and parametric approaches(SPA)for arbitrary nonuniform arrays,among which,one is SPA based on array interpolation(AI-SPA)and the other is SPA based on the manifold separation technique(MS-SPA).These two methods both have the similar good performance as the conventional SPA method.3)To deal with coherent targets using the virtual sparse non-uniform linear array(VSNLA)of the a sparse MIMO radar,two extended spatial smoothing methods are proposed.The characteristics of the VSNLA is analyzed carefully at first.To extend the conventional spatial smoothing technique to VSNLAs,an extended spatial smoothing scheme is proposed by taking advantage of the special array structure of the VSNLA.The VSNLA is divided into multiple coprime subarrays with the same array structure and then the extended spatial smoothing methods are performed by averaging the covariance matrices of all coprime subarrays to rebuild the rank of the covariance matrix of the received signal.To obtain the final DOA estimation,the modified root-MUSIC algorithm is applied to improve the DOA estimation performance after spatial smoothing.4)To overcome the spatial aliasing problem for the equivalent sparse uniform linear array(SULA)of a MIMO radar,two combinations of multiple frequencies are proposed to resolve the ambiguity problem.One is the coprime frequencies combination,the other is an arbitrary frequencies combination.The sufficient anti-aliasing conditions for the two combinations are provided.Based on the fact that a real DOA and its ambiguous positions are periodic distribut ed,two fast DOA algorithms are proposed: i)The partial spectral search-based method uses partial seaching to obtain the ambiguous DOA sets,which dramatically reduces the computational complexity ii)The search-free methods needs no spectral searching for the ambiguous DOA sets.Both method separate the real DOAs from the ambiguous ones by using the same effective matching method.Furthermore,a combined root-MUSIC is proposed for the coprime frequencies combination,which takes the multiple equivalent arrays as different sub-arrays of the same filled ULA and uses polynomials of all coprime frequencies to form a new polynomial with unique roots.To extend the combined root-MUSIC to the arbitrary frequencies combination,an extension of combined root-MUSIC based on manifold separation technique is also proposed.