Subject description - B3M35OFD

Summary of Study | Summary of Branches | All Subject Groups | All Subjects | List of Roles | Explanatory Notes               Instructions
B3M35OFD Estimation, filtering and detection
Roles:PV, PO Extent of teaching:2P+2C
Department:13135 Language of teaching:CS
Guarantors:Havlena V. Completion:Z,ZK
Lecturers:Havlena V. Credits:6
Tutors:Hauser J., Havlena V., Tabaček J. Semester:Z

Web page:

https://moodle.fel.cvut.cz/courses/B3M35OFD

Anotation:

This course will cover description of the uncertainty of hidden variables (parameters and state of a dynamic system) using the probability language and methods for their estimation. Based on bayesian problem formulation principles of rational behavior under uncertainty will be analyzed and used to develop algorithms for parameter estimations (ARX models, Gaussian process regression), filtering (Kalman filter) and detection (likelihood ratio theory) . We will demonstrate numerically robust implementation of the algorithms applicable in real life problems for the areas of industrial process control, robotics and avionics.

Study targets:

Ability to solve engineering problems in the area of estimation and filtering, using rigorous theoretical background.

Content:

MS, LMS and ML estimation. Bayesian approach to uncertainty description, model of dynamic system. Identification of ARX model parameters. Tracking of time varying parameters, forgetting, prior information. Numerically robust algorithms for parameter estimation. Gaussian process regression. Stochastic system, probabilistic state definition, Kalman filter. Kalman filter for colored noise, extended Kalman filter. Stochastic dynamic programming, LQ and LQG controller, certainty equivalence principle. Fault detection and isolation methods. Likelihood ratio - theory and applications. Nonlinear estimation - local vs. global approximation. Monte Carlo methods.

Course outlines:

1. Review of basic concepts of statistics
2. MS, LMS and ML estimation
3. Bayesian approach to uncertainty description, model of dynamic system
4. Identification of ARX model parameters
5. Tracking of time varying parameters, forgetting, prior information
6. Numerically robust algorithms for parameter estimation
7. Gaussian process regression
8. Stochastic system, probabilistic state definition, Kalman filter
9. Kalman filter for colored noise, extended Kalman filter
10. Stochastic dynamic programming, LQ and LQG controller, certainty equivalence principle
11. Fault detection and isolation methods
12. Likelihood ratio - theory and applications
13. Nonlinear estimation - local vs. global approximation
14. Monte Carlo methods

Exercises outline:

Individual assigments - implementation of selected algorithms in Matlab, solution of individual technical problems. Deliverables: running algorithm, technical report. Homeworks: theoretical assignments. Deliverables: report.

Literature:

Lewis, F. L., L. Xie, D. Popa: Optimal and Robust Estimation: With an Introduction to Stochastic Control Theory, CRC Press, 2005. ISBN 978-1-4200-0829-6 Simon, D.: Optimal State Estimation: Kalman, H Infinity, and Nonlinear Approaches. Wiley, 2006, ISBN: 978-0-471-70858-2 Lectures - published on WEB/Moodle Assignments-homework - published on WEB/Moodle

Requirements:

Basics of dynamic system theory, probability and statistics.

Note:

Updated 31. 1. 2016, 3+1

Keywords:

Estimation, filtering, stochastic system, state, nonlinear system, likelihood ratio, monte Carlo method.

Subject is included into these academic programs:

Program Branch Role Recommended semester
MPKYR3_2016 Systems and Control PO 3
MPKYR1_2016 Robotics PV 3
MPKYR4_2016 Aerospace Systems PV 3
MPKYR_2021 Common courses PV 3
MPKYR5_2016 Cybernetics and Robotics PV 3
MPKYR2_2016 Sensors and Instrumentation PV 3


Page updated 7.10.2024 15:51:24, semester: Z/2024-5, Z/2025-6, L/2023-4, L/2024-5, Send comments about the content to the Administrators of the Academic Programs Proposal and Realization: I. Halaška (K336), J. Novák (K336)