Kalman & State Estimation Workshop
Description
This two-day course discusses applications and worked examples of Kalman filtering and other advanced state estimation algorithms. The course focuses on Matlab examples and is thus more like a hands-on workshop rather than a lecture-oriented course. Students are expected to bring their own computers to the class with Matlab, but no Matlab toolboxes are used in this workshop. Students will interactively participate in Matlab programming. Matlab scripts written by the instructor will be provided to all students, and additional scripts will be written during class.
What You Will Learn:
- How can I simulate noise with various types of statistical properties?
- How can I make my filter robust to uncertainty, noise, and modeling errors?
- How can I incorporate problem-specific information in my filter?
- How can I decrease the computational effort of my filter?
- How can I make my filter more numerically stable?
- How can I design an adaptive filter?
- How can I estimate unknown inputs?
- How can I estimate system parameters?
- How can I implement unscented Kalman filters?
- How can I implement particle filters?
- How can I optimize my filter gains when model information is not available?
Course Outline:
- Probability theory
- Random number seeds
- White noise and colored noise
- Correlated random variables
- Linear state estimation
- Compensating for modeling errors
- H-infinity filters
- Steady-state filters
- Constrained filters
- Filter divergence
- Multiple-model filters
- Nonlinear state estimation
- Extended Kalman filters
- System parameter estimation
- Evolutionary filter optimization
- Unscented Kalman filters
- Satellite orbit estimation
- Particle filters
- Correntropy filters
Prerequisites:
ATI’s two-day course “Kalman, H-infinity, and Nonlinear Estimation” or equivalent; familiarity with Matlab programming; availability of laptop or notebook computer with Matlab. Students are encouraged to bring a copy of Dr. Simon’s text, Optimal State Estimation, to class, although the text is not required.