Research On Broadband High-order Subspace Technology In Ultra-short Baseline Localization

With the development of marine resources exploration and other fields,the technology of underwater acoustic localization with acoustic waves to locate underwater targets such as AUVs has become one of the hotspots in related research.Faced with the challenges of strong multipath,strong noise and attenuation in the marine environment,the underwater sound source localization method using the ultra-short baseline system is the main method for underwater targets such as AUV because of its smaller size,simple installation and flexible operation.Up to now,the technology of underwater acoustic localization with the ultra-short baseline is divided into the phase estimation algorithm for narrowband signals and the time delay estimation algorithm for wideband signals.However,boths of them are difficult to achieve reliable localization with high-resolution and wide-area.In the face of the dynamic trajectory localization scene of the target,the calculation accuracy is also poor.What’s more,in the actual marine environment,the robustness to color noise is insufficient because they are difficult to utilize the complete statistical information of the signal.At the same time,the phase estimation algorithm requires additional solutions for the phase wrapping problem and so on.The subspace algorithms in the field of DOA estimation have many similarities with the underwater acoustic localization using the ultra-short baseline.The subspace high-resolution spatial spectrum estimation algorithms represented by MUSIC and ESPRIT can obtain highresolution and High-precision position results.In this thesis,based on the improvement for the feature subspace-like MUSIC algorithm,two aspects of ultra-short baseline static localization and dynamic localization are studied.At first,a broadband high-order subspace method is proposed for the ultra-short baseline static localization.Secondly,a subspace method with adaptive extended Kalman filter is further proposed for ultra-short baseline dynamic localization.The algorithm named RSS-Fourth MUSICAL is proposed for the static localization method with ultra-short baselines.Based on the principle of MUSIC algorithm,the spatial spectral function of the algorithm is modified with the direction vector of any array manifold to apply to ultra-short baseline localization in three-dimensional space at first.Secondly,the static localization of the broadband signals is achieved through focusing transformation by the rotating signal subspace algorithm,which could construct a focusing matrix combined with the coherent signal-subspace methods.In addition,for non-Gaussian distributed colored noise signals,The fourth-order cumulant in the higher-order statistics is introduced.And on this basis,combined with the spatial-temporal smoothing,the RSS-Fourth MUSICAL algorithm is proposed.Finally,simulation experiments in different dimensions are used to verify the effectiveness of the broadband processing of the proposed algorithm and the localization effect in the dual-signal source localization scenario.Experimental results show the proposed algorithm can achieve better localization effect under low signal-to-noise ratio,and achieve better localization performance overall.Aiming at the dynamic localization method of ultra-short baseline,the algorithm named ARF MUSICAL is proposed.Firstly,in this thesis,based on the extended Kalman filter theory,combined with Taylor series expansion,local linear features are extracted by Jacobian matrix.Secondly,combined with the adaptive filtering,the forgetting factor is introduced to control the regularization parameters,and to control continuously abnormal disturbance and state model error.What’s more,a residual term is introduced for the state estimation in the update stage Simultaneously,which can satisfy the localization of nonlinear movement trajectory under nonGaussian distribution.Finally,combined with the position results,which are calculated by RSSFourth MUSICAL algorithm as the measurement data of the adaptive extended Kalman filter,the ARF MUSICAL algorithm is proposed.Lastly,the effectiveness of the proposed algorithm for nonlinear dynamic trajectory localization is verified with different noise environments and different motion trajectories through simulation experiments.The experimental results show that the proposed algorithm can obtain better localization accuracy and precision.After research and experimental analysis,the proposed RSS-Fourth MUSICAL algorithm and the ARF MUSICAL algorithm in this thesis can better realize the underwater target localization of ultra-short baseline.