Study On A New Parameter And Condition Joint Estimation Algorithm In Nonlinear System

At present, non-linear signal processing is a concern and a challenging research topic. it plays an attractive prospect automatic control in wireless communications, economic statistics, artificial intelligence, information fusion, and image processing and other fields. It becomes a major hot spot for many scholars to study in the international. When in the nonlinear system has the unknown parameters, system's estimate duty is:According to system's observation data and the prior estimate information, through certain algorithm, estimated needs to estimate, when the system parameters are unknown, we need to simultaneously estimate system's condition and the parameter. Extended Kalman filter is the most common method to solve non-linear signal processing, it can have a very good estimation result for the weakly non-linear systems, however, when facing the stronger non-linear system with non-Gaussian noises, its estimation accuracy will be greatly reduced, and may cause filter divergence.Particle filter method is a new filtering method in recent years, which can effectively solve the non-linear, non-Gaussian problems. The algorithm bases on the Monte Carlo simulation and recursive Bayesian estimation algorithm, through a sequence samples with the weight to describe the posterior probability density, and then use this approximation to calculate the probability density of the state estimation. From the particle filter was first proposed, it has already been successfully applied to signal processing, target tracking, biometrics and many other fields.In this paper, we in-depth discuss the specific process of the Kalman filter algorithm (KF), extended Kalman filter (EKF), unscented Kalman filter (UKF) algorithm and simulation to compare their state estimation results; we study the particle filter (PF) algorithm, and aimed at particle degradation and particle depletion respects, we discuss specific improvement algorithms, at last, the simulation comparison shows that the estimation accuracy of PF-UKF-MCMC is higher than several other estimation methods.In this paper, it proposes a new joint estimation algorithm, The algorithm uses particle filter methods, combined with the kernel smoothing contraction method, replaces the traditional use of the Gaussian distribution with the standard beta distribution to fit the posteriori distribution of the unknown parameter of the system, in order to achieve the iteration of the parameter estimation of the nonlinear system. The simulation results show that the algorithm in terms of accuracy and convergence is superior to the dual Kalman state and parameter estimation (DEKF).it can effectively resolve the joint estimation problem of the parameters and status in non-linear system.This article in the parameter and the condition union estimate algorithm's foundation, the parameter and the condition expansion is the two-dimensional vector, conducted the simulation research in view of the observational equation for linear and the misalignment two kind of situations, the experiment had indicated this article algorithm might also obtain the very good utilization in the composite signal separation's question.