Chapter 2. Parameter-Based Kalman Filter Training: Theory and Implementation

  1. Simon Haykin
  1. Gintaras V. Puskorius and
  2. Lee A. Feldkamp

Published Online: 13 MAR 2002

DOI: 10.1002/0471221546.ch2

Kalman Filtering and Neural Networks

Kalman Filtering and Neural Networks

How to Cite

Puskorius, G. V. and Feldkamp, L. A. (2002) Parameter-Based Kalman Filter Training: Theory and Implementation, in Kalman Filtering and Neural Networks (ed S. Haykin), John Wiley & Sons, Inc., New York, USA. doi: 10.1002/0471221546.ch2

Editor Information

  1. Communications Research Laboratory, McMaster University, Hamilton, Ontario, Canada

Author Information

  1. Ford Research Laboratory, Ford Motor Company, 2101 Village Road, Dearborn, MI 48121-2053, USA

Publication History

  1. Published Online: 13 MAR 2002
  2. Published Print: 1 OCT 2001

Book Series:

  1. Wiley Series on Adaptive and Learning Systems for Signal Processing, Communications, and Control

Book Series Editors:

  1. Simon Haykin

Series Editor Information

  1. Communications Research Laboratory, McMaster University, Hamilton, Ontario, Canada

ISBN Information

Print ISBN: 9780471369981

Online ISBN: 9780471221548

SEARCH

Keywords:

  • parameter-based Kalman filter training;
  • theory;
  • implementation;
  • network architectures;
  • EKF procedure;
  • decoupled EKF (DEKF);
  • multistream training;
  • computational considerations;
  • extensions;
  • enhancements

Summary

This chapter presents a brief discussion of the types of feedforward and recurrent network architectures that are to be considered for training by extended Kalman filter (EKF) methods. A discussion of the global EKF training method, followed by recommendations for setting of parameters for EKF methods, including the relationship of the choice of learning rate to the initialization of the error covariance matrix is included. Treatments of the decoupled extended Kalman filter (DEKF) method are provided. An extensive discussion is given on a variety of issues relating to computer implementation, including derivative calculations, computationally efficient formulations, methods for avoiding matrix inversions, and square-root filtering for computational stability. An overview is provided of applications of EKF methods to a series of problems in control, diagnosis, and modeling of automotive powertrain systems. The chapter concludes with a discussion of the virtues and limitations of EKF training methods, and provides a series of guidelines for implementation and use.