The work presented in this report derives a sequential method for on-line estimation of the state variables and parameters of discrete, non-linear, dynamical systems. A discrete version of Pontryagin's Maximum Principle is employed to obtain the canonic equations of the least-squares optimal estimator. A discretized invariant imbedding technique is then applied to solve the resulting two-point boundary value problem. Finally, a system of sequential equations is obtained by application of variational methods to the optimal trajectory. The result is a sequential estimation scheme conceptually related to existing methods developed for continuous systems. The method presented has the advantage of direct applicability to discrete systems and provides for the inclusion of higher-order terms not usually considered by other methods. As a result of these inherent features, the process has been found to provide a faster, more stable estimate of the system variables. In addition, a minimum of a-priori statistical information is required. (Author).

The problem of state and parameter estimation for noisy discrete-time nonlinear dynamic systems is examined from the viewpoint of marginal maximum likelihood estimation. Approximate algorithms for sequential prediction, filtering, and smoothing are developed. The former two are in agreement with previous results; the latter is new. A technique for iterative-sequential filtering and smoothing using Newton's method is indicated. A numerical example is included to illustrate the results. (Author).

In these notes, we introduce particle filtering as a recursive importance sampling method that approximates the minimum-mean-square-error (MMSE) estimate of a sequence of hidden state vectors in scenarios where the joint probability distribution of the states and the observations is non-Gaussian and, therefore, closed-form analytical expressions for the MMSE estimate are generally unavailable. We begin the notes with a review of Bayesian approaches to static (i.e., time-invariant) parameter estimation. In the sequel, we describe the solution to the problem of sequential state estimation in linear, Gaussian dynamic models, which corresponds to the well-known Kalman (or Kalman-Bucy) filter. Finally, we move to the general nonlinear, non-Gaussian stochastic filtering problem and present particle filtering as a sequential Monte Carlo approach to solve that problem in a statistically optimal way. We review several techniques to improve the performance of particle filters, including importance function optimization, particle resampling, Markov Chain Monte Carlo move steps, auxiliary particle filtering, and regularized particle filtering. We also discuss Rao-Blackwellized particle filtering as a technique that is particularly well-suited for many relevant applications such as fault detection and inertial navigation. Finally, we conclude the notes with a discussion on the emerging topic of distributed particle filtering using multiple processors located at remote nodes in a sensor network. Throughout the notes, we often assume a more general framework than in most introductory textbooks by allowing either the observation model or the hidden state dynamic model to include unknown parameters. In a fully Bayesian fashion, we treat those unknown parameters also as random variables. Using suitable dynamic conjugate priors, that approach can be applied then to perform joint state and parameter estimation. Table of Contents: Introduction / Bayesian Estimation of Static Vectors / The Stochastic Filtering Problem / Sequential Monte Carlo Methods / Sampling/Importance Resampling (SIR) Filter / Importance Function Selection / Markov Chain Monte Carlo Move Step / Rao-Blackwellized Particle Filters / Auxiliary Particle Filter / Regularized Particle Filters / Cooperative Filtering with Multiple Observers / Application Examples / Summary

Monte Carlo methods are revolutionizing the on-line analysis of data in many fileds. They have made it possible to solve numerically many complex, non-standard problems that were previously intractable. This book presents the first comprehensive treatment of these techniques.

In these notes, we introduce particle filtering as a recursive importance sampling method that approximates the minimum-mean-square-error (MMSE) estimate of a sequence of hidden state vectors in scenarios where the joint probability distribution of the states and the observations is non-Gaussian and, therefore, closed-form analytical expressions for the MMSE estimate are generally unavailable. We begin the notes with a review of Bayesian approaches to static (i.e., time-invariant) parameter estimation. In the sequel, we describe the solution to the problem of sequential state estimation in linear, Gaussian dynamic models, which corresponds to the well-known Kalman (or Kalman-Bucy) filter. Finally, we move to the general nonlinear, non-Gaussian stochastic filtering problem and present particle filtering as a sequential Monte Carlo approach to solve that problem in a statistically optimal way. We review several techniques to improve the performance of particle filters, including importance function optimization, particle resampling, Markov Chain Monte Carlo move steps, auxiliary particle filtering, and regularized particle filtering. We also discuss Rao-Blackwellized particle filtering as a technique that is particularly well-suited for many relevant applications such as fault detection and inertial navigation. Finally, we conclude the notes with a discussion on the emerging topic of distributed particle filtering using multiple processors located at remote nodes in a sensor network. Throughout the notes, we often assume a more general framework than in most introductory textbooks by allowing either the observation model or the hidden state dynamic model to include unknown parameters. In a fully Bayesian fashion, we treat those unknown parameters also as random variables. Using suitable dynamic conjugate priors, that approach can be applied then to perform joint state and parameter estimation.

Classification, Parameter Estimation and State Estimation is a practical guide for data analysts and designers of measurement systems and postgraduates students that are interested in advanced measurement systems using MATLAB. 'Prtools' is a powerful MATLAB toolbox for pattern recognition and is written and owned by one of the co-authors, B. Duin of the Delft University of Technology. After an introductory chapter, the book provides the theoretical construction for classification, estimation and state estimation. The book also deals with the skills required to bring the theoretical concepts to practical systems, and how to evaluate these systems. Together with the many examples in the chapters, the book is accompanied by a MATLAB toolbox for pattern recognition and classification. The appendix provides the necessary documentation for this toolbox as well as an overview of the most useful functions from these toolboxes. With its integrated and unified approach to classification, parameter estimation and state estimation, this book is a suitable practical supplement in existing university courses in pattern classification, optimal estimation and data analysis. Covers all contemporary main methods for classification and estimation. Integrated approach to classification, parameter estimation and state estimation Highlights the practical deployment of theoretical issues. Provides a concise and practical approach supported by MATLAB toolbox. Offers exercises at the end of each chapter and numerous worked out examples. PRtools toolbox (MATLAB) and code of worked out examples available from the internet Many examples showing implementations in MATLAB Enables students to practice their skills using a MATLAB environment