Vous êtes sur la page 1sur 19

# MECH468 : Modern Control Engineering

## MECH522 : Foundations in Control Engineering

L14 : Kalman Decomposition
Dr. Ryozo Nagamune
Department of Mechanical Engineering
University of British Columbia

MECH 468/522

## Review & Todays topic

So far, we learned controllability and observability:
Condition
Decomposition
Duality

## Today, we will study the combination of

decompositions for controllability and observability,
called Kalman decomposition.

MECH 468/522

## Rudolf Emil Kalman (1930-)

Hungarian-born American
Electrical engineer & mathematician
During 1950s & 60s, he developed
state-space control theory
Controllability, observability
Kalman filter

## Most of the theory in this course

have been established by Kalman!
MECH 468/522

Two decompositions

System
Decomposition for
controllability

Controllable part
Uncontrollable part

Decomposition for
observability
Observable part
Unobservable part

MECH 468/522

System

## Controllable & observable part

Controllable & unobservable part
Uncontrollable & observable part
Uncontrollable & unobservable part
MECH 468/522

Kalman decomposition
Conceptual figure (Not block-diagram)
System

MECH 468/522

Kalman decomposition
Every SS model can be transformed by z=Tx for
some appropriate T into a canonical form:

## Note the decomposition structure for controllability.

MECH 468/522

Kalman decomposition 2
If the second and the third state vectors are
exchanged, one can get another form:

## Note the decomposition structure for observability.

MECH 468/522

Remark
It may happen that some states are missing.
Ex. No controllable and unobservable part

MECH 468/522

Remarks
(Aco,Bco): controllable & (Aco,Cco): observable
Transfer function is determined by ONLY
controllable & observable parts.

## If uncontrollable and/or unobservable parts are

unstable, we need to change the structure
(actuators/sensors) of the system, because no
output feedback control can stabilize the system.
(next slide)
MECH 468/522

10

Eigenvalues of A-matrix

## These have to be stable

for output feedback control.
MECH 468/522

11

## How to find T? (review)

For controllability, we used image space of
controllability matrix.

## For observability, we used kernel space of

observability matrix.

MECH 468/522

12

decomposition?

MECH 468/522

13

## Very simple example: revisited

Three cars with one input and two outputs

Car1

Car2

Car3

x1: position
x2: velocity
x3: position
x4: velocity

## Guess which state is

controllable and/or observable!

x5: position
x6: velocity
MECH 468/522

14

## SS model of three car example

Velocity of
1st mass

Position of
3rd mass
MECH 468/522

15

Derivation of T
Controllability matrix

Observability matrix

## (Inverse of T)=[e2,e1,{e5,e6},{e3,e4}](=T in this case)

MECH 468/522

16

Example (contd)
After transformation

MECH 468/522

17

## Remark: Matlab will return matrices corresponding the

states in an order different from the order in Slide 7.
MECH 468/522

18

Summary
Kalman decomposition
Combination of
Decomposition for controllability
Decomposition for observability

## How to find coordinate transformation matrix T

Image space of controllability matrix
Kernel space of observability matrix

MECH 468/522

19