AA/EE/ME 549: Estimation and System Identification Spring 2007 T/Th 12:30-1:50, Loew 202

Lectures: T/Th 12:30-1:50am, Loew Hall 202
Homework section (optional):  TBD

Course Description

A great many control design and analysis applications involve systems that are not well-understood, and for which detailed models are unavailable. Without an estimate of the state variables of a system, standard control theoretic techniques cannot be applied. To address this problem, system observers have been developed for a number of classes of systems.

This course will focus on development of observers and optimal observers for both discrete and continuous time with emphasis on continuous time. Both linear and nonlinear systems will be considered. The course will include a project - with students working individually or in pairs.

The goal of this course is to enable all students to have the skills and knowledge to successfully apply estimation techniques to a variety of applications.

Prerequisites: EE 505, AMATH 506 or STAT 506; Recommended AA/EE/ME 548

Topics:

1. Dynamical Systems
2. Least Squares
3. Probability and Random Variables
4. Parameter Estimation and System Identification
5. Kalman Filtering (standard, unscented, scented, extended)
6. Smoothing
7. Particle Filters

Expectations

My role:  The role of an instructor is to help the students acquire new knowledge and skills more quickly than they could on their own, to guide the approach to learning with effective tools, to provide completeness of subject matter, and to place material in context relative to the larger field.

Student role:  The role of a student is, of course, to learn.  Students in this course are expected to read the notes associated with a topic before the material is presented in class, to prepare questions on the reading (need for clarification, connections to previous material, placement of the material in a larger context, etc), to not wait until 24 hours before assignments are due to begin them, to utilize the office hours of the professor and TA, and to interact professionally with all members of the course.

General:  Students interested in pursuing graduate degrees in control theory and robotics come from a variety of backgrounds.  From the point of view of course material, this course provides the fundamental groundwork for estimation theory.  Students who take this course generally find that the material is challenging, that homework requires significant effort, but that in the end, the time and effort are well worth the payoff.   

Office hours:  All students are requested to attend office hours once during the first week of class.  This request allows me a chance to get to know you personally, and familiarizes you with the path to at least one of my offices. 

EDGE:  All EDGE students are requested to send me a phone number and time when I can call you during the first week of classes.  My experience with distance learning in the past has been that without early efforts to create an active interaction between faculty and students, inertia may set in. 

Resources

·        Robotics, Control and Mechatronics web site:  http://www.engr.washington.edu/rcm

·        Class mailing list:  All enrolled students can send messages to the class, TA and instructor through the mailing list:  TBA.

Homework and Exam Policy

Collaboration on homework assignments is allowed. You may consult outside reference materials, other students, or the instructor. All solutions that are handed in should reflect your understanding of the subject matter at the time of writing. No collaboration is allowed on the midterm or the final exam.

Homework format:  When composing your homework for submission, please adhere to the following guidelines:  (1) all problems should be submitted in the same order as in the assignment, (2) give each problem the same label as in the assignment, (3) begin each problem by restating the problem then indicate how you will approach the solution, (4) show all relevant work indicating how you reach your solution, (5) keep relevant information associated (e.g. eigenvectors belong to specific eigenvalues), (6) clearly mark your answer, (7) indicate or discuss why your answer is correct or appropriate (e.g. check your answer), and (8) take pride in your work—neatness counts in whatever profession you have in the future, so practice now!  For more information see the Five Steps to Problem Solving.

Homework section:  Homework is due at the beginning of lecture on Tuesdays (12:30pm).  EDGE students should email a copy of their assignment to Prof. Morgansen so that it arrives before 12:30pm on Tuesdays.  An optional homework section will take place on Fridays 1:30-2:30 in Loew 202.  In this section, students will take turns presenting their solutions to the homework assignment that was due that day.  Participation is not obligatory and will provide 5 pts extra credit to your lowest homework score (list of presentation dates).  The purpose of this section and the approach taken in it is to give students practice with presentation abilities, to practice problem solving in front of an audience (part of all qualifying exams), and to answer all questions about an assignment before starting the next one.

Homework section format:  Assignment list.  Three people will present during each section (following alphabetical order of the entire class).  You are welcome to exchange slots, but please let me know if you choose to do so or if you do not wish to participate.  Each presentation will earn 5pts of extra credit on a given assignment (assignments will generally have a total of 30pts).  If the assignment does not have six problems, it will be split into sections (as indicated on the list).  Each presenter will have 20 minutes total including time to put on the microphone.  Plan accordingly.  Be aware of the following guidelines for presenting:  plan for the time limit, start at the top of one board section, work down the section, then start in the next board section (partly this is for video taping, but also good style in general), state your reasoning clearly, and talk through the solution.  Credit is given for presentation and a valid attempt at solution, not necessarily an entirely correct solution.

Grading

1. Homework: 30%. There will be 9 homework assignments due at 12:30pm on Tuesdays. Each problem set will have about 6 problems and each problem is worth 5 points. Late homework will not be accepted without prior permission from the instructor.  Homework must be written up following the five steps of problem solving.
2. Midterm: 20%. The midterm will be a take-home exam posted on April 24 and due on May 1 at 12:30pm.
3. Final: 30%. The final exam will be a take-home exam posted on June 1 and due on June 8 at 5pm.
4. Project: 20%. Students will complete a project of designing an observer for a physical system. For your project, you may choose any system (real or simulated), and build an estimator that reconstructs the system state from observations.  The system may be anything as long as it is a real system (e.g. tracking live fish, tracking robotic fish, estimating the state of polymer chains, tracking vortices in fluid, estimating position and orientation of an aircraft).  Data or simulators for a number of systems can be obtained from various faculty projects.  Project plans are due April 24. The project plan should outline your project according to the five steps of problem solving.  The final project must be typed.  Project guidelines.

Textbook (required)

J. L. Crassidis and J. L. Junkins, "Optimal Estimation of Dynamic Systems," Chapman & Hall/CRC, 2004.

References (on reserve in Engineering Library)

1. A. Gelb, "Applied Optimal Estimation," MIT Press, 1974.

This schedule is an initial guideline and is subject to adjustment as the course progresses.

EDGE lecture videos: http://www.engr.washington.edu/edge/aa549/aa549vd.html

Prof. Morgansen will be traveling the following dates (lectures marked with a * will be pre- or post-recorded):
April 12 -> Thursday April 12, 1:30-3pm Loew 216
May 14-17-> Monday May 14, 9-10:30am Loew 202 and Friday May 18, 3:30-5pm Loew 202
May 31-> Wednesday May 30, 11:30-1pm Loew 202

Schedule

Date	Topics	Reading	Assignments
Mar 27	Linear and Nonlinear Dynamical Systems	3.1-3.6, linearization notes	Homework #1 Assignment (Solutions)
Mar 29	Rigid Body Dynamics	3.7-3.11, ODE notes
Apr 3	Linear Least Squares	1.1-1.3	Homework #2 Assignment (Solutions)
Apr 5	Nonlinear Least Squares and Basis Functions	1.4-1.7
Apr 10	Probability and random processes	online notes (Gelb 2.2)	Homework #3 Assignment (Solutions, Matlab code)
Apr 12*	Minimum Variance	2.1-2.2
Apr 17	Maximum Likelihood	2.3-2.5	Homework #4 Assignment (Solutions)
Apr 19	Bayesian Estimation	2.6-2.8
Apr 24	GPS and Attitude Determination Orbit Determination and Aircraft Parameter Identification	4.1-4.2 4.3-4.6	Midterm (Solutions)
Apr 26	Discrete Time Kalman Filter	5.1-5.3, fig51.m, true51.m
May 1	Continuous Time Kalman Filter	5.4-5.5	Homework #5 Assignment (Solutions)
May 3	Extended Kalman Filter and Colored Noise	5.6
May 8	Unscented Kalman Filter	5.7-5.8	Homework #6 Assignment (Solutions)
May 10	Discrete and Continuous Fixed Interval Smoothing	6.1.1-6.1.2
May 15*	Nonlinear Fixed Interval Smoothing	6.1.3
May 17*	Discrete and Continuous Fixed Point Smoothing	6.2.1-6.2.2
May 22	Discrete and Continuous Fixed Lag Smoothing	6.3.1-6.3.2	Homework #7 Assignment (Solutions)
May 24	Histogram Filters	online notes
May 29	Particle Filters	online notes
May 31*	Applications and Exampls	EE549_HW7_5_49.m, polfun.m, polfun_id.m Fish tracker UKF:  ukf.m, writeupFinal.pdf Particle filter code: p5_10.m, p5_10ode.m
	FINAL EXAMINATION