Model Building For Dynamic Systems
525.786

Course Description

The course presents the theory and practice of system identification, which is the process of estimating statistical system models from data. State-space methods are used in discussing both adaptive filtering and control, where the system model must be estimated on-line, and off-line model building, for simulation development and validation, and test and evaluation. Practical implementations are covered along with example applications to real-world problems. Methods include: nonlinear extended Kalman, residual analysis, and multiple model adaptive filters; the eigensystem realization algorithm; canonical variate analysis; and subspace model identification, as well as simultaneous perturbation stochastic approximation and maximum likelihood, prediction error minimization, and model structure estimation methods. Applications include adaptive control of a farm vehicle, urban traffic control, missile inertial guidance modeling, and GPS receiver modeling.

Syllabus

1. Review of Kalman Filtering
2. System Identification (SID) Basics
3. Nonlinear Extended Kalman Filters as Adaptive On-Line Filters
4. Residual Analysis and Multiple Model Adaptive Filters
5. Observer/Kalman Filter Identification
6. Adaptive Control of a Vehicle Using GPS
7. Subspace Identification and Control
8. Maximum Likelihood State-Space SID(part 1)
9. Maximum Likelihood State-Space SID(part2)
10. Simultaneous Perturbation Stochastic Approximation (SPSA)
11. Missile Inertial Guidance Modeling with GPS
12. GPS Receiver Modeling
13. Simulation Model Validation, Verification, and Certification
14. Review of Class Projects

Prerequisites
525.745 Applied Kalman Filtering or equivalents.

Instructor
Larry Levy joined APL in 1971, working on performance analysis of Kalman-filter-integrated transit/inertial navigation systems in the space department. He was appointed to the Principal Professional Staff in 1978. Dr. Levy has had a series of increasingly important responsibilities in the areas of navigation system development and system test and evaluation. He is recognized as an expert in Kalman filtering and system identification and is the author of numerous papers and presentations on these subjects. He was the co-developer of the GPS translator concept in SATRACK and was instrumental in developing the end-to-end methodology for Trident II accuracy evaluation. Dr. Levy is also a graduate lecturer in Kalman filtering and system identification courses in the Johns Hopkins University Whiting School of Engineering and a short course lecturer on these subjects for Navtech Seminars.
E-mail the instructor.

Computer Lab Requirements
PC with MATLAB is required for homework.

Textbook
none available

Return to Electrical Engineering Courses | Electrical Engineering | Part-Time Engineering

June 1998