Lecture 28 - CCAR - University of Colorado Boulder

```ASEN 5070: Statistical Orbit Determination I
Fall 2014
Professor Brandon A. Jones
Lecture 28: Orthogonal Transformations
Boulder

Lecture quiz due at 5pm

Exam 2 – Friday, November 7
Boulder
2











Minimum Variance
Conventional Kalman Filter
Extended Kalman Filter
Prediction Residuals
Handling Observation Biases
Numeric Considerations in the Kalman
Batch vs. CKF vs. EKF
Effects of Uncertainties on Estimation
Potter Square-Root Filter
Cholesky Decomposition w/ Forward and
Backward substitution
Singular Value Decomposition Methods
Boulder
3
Least Squares via Orthogonal Transformations
Boulder
4
Boulder
5


Recall the least squares cost function:
By property 4 on the previous slide and Q an orthogonal
matrix:
Boulder
6
Boulder
7
Boulder
8

The method for selecting R defines a
particular algorithm
◦ Givens Transformations (Section 5.4)
◦ Householder Transformation (Section 5.5)
◦ Gram-Schmidt Orthogonalization
 Not in the book and we won’t cover it
Boulder
9
LS Solution via Givens Transformations
Boulder
10
Boulder
11
Boulder
12

Consider the desired result

To achieve this, we select the Givens matrix such that

We then use this transformation in the top equation
Boulder
13

We do not want to add non-zero terms to the
previously altered rows, so we use the
identity matrix except in the rows of interest:
Boulder
14

After applying the transformation, we get:

Repeat for all remaining non-zero elements
in the third column

What if the term is already 0 ?
Boulder
15

Need to find the orthogonal matrix Q to yield a
matrix of the form of the RHS

Q is generated using a series of Givens
transformations G
Boulder
16
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
17
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
18
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
19
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
20
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
21
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
22
We select G to get a zero for the term in red:
To achieve this, we use:
Boulder
23

We now have the required Q matrix (for this
conceptual example):
Boulder
24
Boulder
25
Givens Transformations – An New Example
Boulder
26

Consider the case where:
Boulder
27
Boulder
28
Boulder
29
Boulder
30

We then have the matrices needed to solve
the system:
Boulder
31
Batch vs. Givens
Boulder
32

Consider the case where:

The exact solution is:

After truncation:
Boulder
33



Well, the Batch can’t handle it. What about
Cholesky decomposition?
Darn, that’s singular too.
Let’s give Givens a shot!
Boulder
34
Boulder
35
Boulder
36
Boulder
37

Hence, Givens transformations give us a solution
for the state
Home Exercise: Why is this true?
Note: R is not equal to H !
Boulder
Still a problem w/ P !
38


Givens uses a sequence of rotations to
generate the R matrix