Fundamentals of stochastic filtering (Q930340): Difference between revisions

Stochastic filtering deals with the problem of estimating an evolving dynamical system customarily modelled by a stochastic process, the signal process. The signal process cannot be observed directly. Instead a second process, the observation process, is known. This observation process is assumed to be a function of the signal and some noise. Mathematically, the inference about the signal means estimation of the conditional distribution of the signal given the observation process. This book provides a rigorous mathematical treatment of the nonlinear stochastic filtering problem with particular emphasis on numerical methods. The text is divided into two parts. The first part lays the theoretical foundations for the topic. It starts by proving that the conditional distribution of the signal can be viewed as a stochastic process with values in the space of probability measures. Then evolution equations are derived and two particular examples, the diffusion case and Markov processes with finite state space, illustrate the theory. Finally, finite-dimensional filters are discussed, such as the Kalman-Bucy filter. The second part focuses on numerical algorithms to solve a given filtering problem. The authors give an overview of existing methods, such as the extended Kalman filter, approximations by finite-dimensional filters, spectral methods and particle filters. Then a detailed discussion of particle methods in continuous as well discrete time are given. The text is essentially self-contained, only some probability theory and basics of stochastic processes are assumed. In an appendice the required results from measure theory and stochastic analysis are stated and proved. Intended readers are researchers and graduate students that have an interest in theoretical aspects of stochastic filtering. The text is supplemented with many exercises and detailed solutions. The authors have taken much care to present the arguments as completely and carefully as possible within the confines of a graduate-level text. The reader is guided through the theory by detailed motivations of the main concepts that make it much easier to understand and appreciate the theory. To sum up, one can say that this book will certainly become a standard reference for teaching and working in the field of stochastic filtering.