Nonlinear nonnormal dynamic models: State estimation and software
Consider a discrete-time nonlinear nonnormal system \[ x_{k+ 1}= f_k(x_k)+ w_k,\quad z_k= h_k(x_k)+ v_k, \] where \(x_k\), \(z_k\) represent state and measurement vectors respectively, \(f_k\), \(h_k\) are known vector functions, and \(w_k\), \(v_k\) are mutually uncorrelated white noise sequences. The initial a priori density of the state vector, state and measurement noise are represented in the form of a discrete normal density mixture \[ p(c)= \sum^\ell_{i=1} p(c| s_i)\cdot Pr(s_i). \] Here, \(Pr(s_i)\) is the probability of a phenomenon \(s_i\) \(p(c| s_i)\) is the conditional normal density function of random value \(c\). The a posteriori density of the state vector has the same form. Using a suitable linearization, the filtering density function \(p(x_k| z_0,z_1,\dots, z_k)\) and the predictive density function \(p(x_{k+1}| z_0,z_1,\dots, z_k)\) are presented. The explanation is given only for \(k= 0\); for \(k= 1,2,\dots\) the algorithm is only formulated. A software package based on MATLAB and on the Symbolic Math Toolbox was developed for solving the described recursive Bayesian nonlinear filtering problems. This tool can be used for tracking time-varying parameters, fault detection, hypothesis testing, robustification of filters due to outliers. For the entire collection see [Zbl 0892.00024].
- scientific article; zbMATH DE number 1053273
- Sequential analysis of nonlinear dynamic systems using particles and mixtures
- Filtering discrete time nonlinear systems with unknown parameters: a nonparametric approach.
- scientific article; zbMATH DE number 4088785
- Nonparametric estimation of a posteriori distributions in nonlinear filtering
This page was built for publication: Nonlinear nonnormal dynamic models: State estimation and software
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4400825)
