Study On Parameter Estimation And Intelligent Cognitive Architecture Of MIMO Radar

Compared with the phased array radar,the multiple input and multiple output(MIMO)radar exploits waveform diversity,polarization diversity,frequency diversity,spatial diversity and other technologies,and achives the advantages of wider virtual aperture,greater freedom,stronger target recognition ability,and better anti-jamming performance.Therefore,MIMO radar has a high research significance in the field of radar.According to the distance between antennas,MIMO radar can be divided into centralized MIMO radar and distributed MIMO radar.The distance between antenna units of centralized MIMO radar is usually very close,roughly equivalent to the wavelength.In this paper,the direction of arrival(DOA)parameters are studied with high resolution by means of atomic norm?SBL and Kalman filter,and the architecture of intelligent cognitive radar is studied.The specific research content mainly includes the following three aspects:1.For the single snapshot DOA estimation of the centralized MIMO radar,traditional methods based on non-sparse reconstruction,such as subspace method multiple signal classification(MUSIC),usually need multiple snapshots to obtain a reasonable covariance matrix,while common methods based on sparse reconstruction,such as sparse Bayesian learning and so on usually have high computational complexity.In this paper,an atom norm denoising method is proposed to obtain the sparse approximation of the received signal in MIMO radar system,in which the sparsity is measured by atom norm.However,this kind of atomic norm denoising problem is a nonconvex optimization problem,which can not be solved effectively.In this paper,a semi positive definite matrix is constructed by deriving the dual norm of the atomic norm.The problem of denoising the atomic norm is transformed into a semi-positive definite programming problem of its dual norm.The semi positive programming problem is a convex optimization problem,which can be effectively solved to remove the noise in the originally received signal.Finally,the DOA is estimated by searching the peak value of spatial spectrum.The simulation results show that the proposed method achieves the ultra-high resolution DOA estimation results with only one snapshot for fast moving targets in MIMO radar system.2.Based on the problem of DOA tracking estimation for centralized MIMO radar,a high-resolution DOA tracking estimation method combining sparse Bayesian learning method and Kalman filtering method is proposed.Firstly,the DOA estimation method based on Sparse Bayesian learning(discretized grid,sparse approximation of scattering coefficient of each grid,probability distribution function by calculating its mean value and variance,non-zero scattering coefficient represents that the grid is the target)is used to get the initial estimation result of the target.Then,according to the time correlation of the moving target,the Kalman filter is used for fast moving target,The angle of moving target is processed by continuous tracking and filtering,and the high-resolution DOA estimation of target in tracking mode is obtained.The simulation results show that the method can get high-resolution DOA estimation of fast moving target tracking and solve the off-grid problem.3.With the rapid development of artificial intelligence technology and modern war cognitive technology,the old cognitive radar(CR)architecture can not meet the needs of cognitive radar development.Based on this,a new architecture of intelligent cognitive MIMO radar is proposed.The feedback from the receiver to the reflector is extended to the feedback from the receiver to the signal processing and then to the transmitter.An intelligent environment sensing method based on reinforcement learning OOEDA method is proposed.In combination with the off-line training and on-line processing of artificial intelligence,A new signal processing method in which deep learning signal processing is integrated with traditional signal processing is proposed,this new architecture use big data analysis technology to mine radar data,unify radar architecture and internal and external interface standards.The feasibility of the new intelligent cognitive MIMO radar architecture is verified by the designed digital / semi physical / physical simulation test platform.The adaptive optimization and DOA estimation of the beam using reinforcement learning DDPG(Deep Deterministic Policy Gradient)network verify that the new signal processing method combined with artificial intelligence method is better than the traditional signal processing method in terms of performance and intelligent cognition.The architecture is based on cognitive radio idea and artificial intelligence technology,and can be software defined.