State Estimation Method For Correlated Noise Systems

In the target tracking,due to the particularity of the tracking model or the special data processing method,the measurement noise is no longer the independent Gaussian white noise that is traditionally considered,but the measurement noise is related to the measurement matrix and the measurement noise.Special target tracking issues such as autocorrelation.In order to deal with the target tracking problem under the nonindependent non-white noise measurement noise,we deduct the new target tracking method under the basic MMSE estimation framework,and use more prior information to improve the target.Tracking accuracy,the main work of the paper is as follows:(1)A new pure angle target tracking algorithm-quasi-linear-MMSE.Due to the particularity of the pure angle tracking problem,the azimuth measurement and the target state vector are nonlinear functions.Therefore,a pseudo-measurement is first constructed by mathematical operations,and the nonlinear measurement equation is quasi-linearized.This step leads to the correlation between the new measurement matrix and the measurement noise.Therefore,the traditional algorithm PLKF has serious deviation problems,making PLKF Tracking performance is greatly reduced,so it needs to be improved.The previous improved algorithm for PLKF is based on the compensation of PLKF deviation.These methods are not only limited in motion model,but also the theoretical framework is incomplete and unified.Therefore,a new pure angle target tracking algorithm,PL-MMSE,is proposed.This method is a new derivation under the unified MMSE estimation framework,which guarantees the theoretical integrity and low computational complexity of the new algorithm.And because PL-MMSE implicitly deals with the correlation between measurement matrix and measurement noise in the MMSE estimation framework,a major shortcoming of the pure angle target tracking problem is that the deviation problem will be improved.By comparing the simulation results of different pure-angle target tracking algorithms,the new method,PL-MMSE,is improved in terms of position and velocity RMSE and Bnorm performance,consistency test comparison and algorithmic computational complexity performance compared with other methods.(2)An interactive multi-model algorithm based on pure-angle target tracking.When the pure angle tracking algorithm using a single fixed motion model is used to track the maneuvering target,since the motion model is set a priori,the moving target cannot be tracked well,so its tracking effect is generally not good,especially It is the increasing maneuverability of the current moving target,which makes it difficult for the tracking algorithm of a single fixed model to identify the moving target motion parameters in time and accurately,thus causing the performance of the tracking algorithm to be degraded due to the inaccuracy of the motion model.In this case,the classic algorithm of maneuvering target tracking,the IMM estimation algorithm,is used to track the maneuvering target.The algorithm has a model transition probability matrix obeying the Markov chain,and considers the interaction between multiple motion models to obtain the state estimation of the target.It is considered to be the most effective maneuvering target tracking method,taking into account the advantages of the algorithm to estimate performance and reduce the amount of calculation.Multi-module parallel computing greatly increases computational efficiency.The proposed pure angle tracking algorithm,PL-MMSE,is applied to the IMM estimation algorithm to track the maneuvering target.The simulation results show that the PL-MMSE exhibits good position and velocity RMSE and BNorm tracking performance when tracking maneuvering targets.(3)A new state estimation algorithm for correlated measurement noise.In the early data processing,in order to improve the data rate,the sliding window is used to accumulate time to achieve the purpose of enhancing the signal-to-noise ratio.Since it is the same noise source,there is a problem of measuring the noise autocorrelation.A new state estimation algorithm for correlated noise measurement-measurement correlation-MMSE.The new MC-MMSE is a new derivation under the unified MMSE framework,which mainly includes the three steps of measurement conversion,correlation measurement noise covariance calculation,and solving new state estimation and covariance matrix estimation.Deducing the theoretical integrity of the process.The initialization of MC-MMSE algorithm adopts three-point initialization method.Compared with CPMKF algorithm simulation,MC-MMSE improves the RMSE performance of position and speed under the condition of large noise correlation coefficient.