1

State-space analysis 2
controllability
J A Rossiter

Slides by Anthony Rossiter

 2

Introduction
• The first video looked at definitions of stability,
which is the inherent boundedness/convergence
properties of a system response, with a bounded
input.
x  Ax  Bu; y  Cx  Du
• However, in general we are interested not just in
convergence, but also the asymptotic value being
the one we desire.
• Controllability links to the ability to achieve the
desired asymptotic state.
Slides by Anthony Rossiter
 3

Controllability (reachability)
A system is controllable if it is possible to
determine an appropriate input signal, for any
initial state, which will achieve a specified final
state at a specified time.

x(0) u(t)

An equivalent definition
exists for discrete time
but needs at least n x(tend)
samples.
Slides by Anthony Rossiter
 4

Remark on Controllability
A system given in control canonical form is always
fully controllable.
 an 1  an  2   a1  a0   
 1  0  1
0  0 0
d
z     z   u;
dt         
  0
0 0  1 0  
   B
A

y  bn 1 bn  2  b0 z
 
C

Slides by Anthony Rossiter

 5

Concepts of range
Let a vector x of dimension n be made up of a
linear combination of ‘n’ other vectors wi.
x  1w1   2 w2   n wn
Does there exist choices of αi such that x can take
any value in n-dimensional space?
The answer is yes, if and only if the matrix W is full
rank.   1
W  w1 w2  wn    1
    W x
 n 
Slides by Anthony Rossiter
 6

Controllability
A system is controllable if it is possible to
determine an appropriate input signal which will
achieve a specified final state x(t).
If x(t) can be represented as:
x(t )  1w1   2 w2   n wn
W  w1 w2  wn 
Then controllability reduces to checking the rank of
matrix W and the ensuring one can choose the
parameters αi freely.
Slides by Anthony Rossiter
 7

Using eigenvalue/vector decomposition

This decomposition is useful for identifying
whether there are regions in the state space which
cannot be reached, from an arbitrary start point.
Without out loss of generality, we will assume the
initial condition is zero!
t t
x  e A ( t  )
Bud    w e i ( t  ) T
v Bu ( )d
i
i  i
0 0  iT

W  w1 w2  wn  W is full rank.

Slides by Anthony Rossiter
 8

Using eigenvalue/vector decomposition

It is clear that x(t) has the structure indicated a few
slides ago, that is:
t
x(t )   wi  e i ( t  ) T
v Bu ( )d
i
 i
0  iT

x(t )   wi i  i   e i ( t  ) T
v Bu ( )d
i
i

It is clear that to choose x(t) arbitrarily 2 conditions are

required:
• αi≠0, and indeed can be specified as required.
• W is full rank (already known).
Slides by Anthony Rossiter
 9

Using eigenvalue/vector decomposition

We require that αi can be chosen freely.

 i  vB e T
i
i ( t  )
u ( )d x   wi i
i
i

Assuming that input u(t) is a free variable so that one

can manage the output of the integration, then αi can
be chosen freely iff βi is non-zero.
It is clear that βi=0 αi=0, so a final state with a
component in the corresponding eigenvector direction
cannot be reached!
Slides by Anthony Rossiter
 10

Using eigenvalue/vector decomposition

A system is fully controllable if and only if the VB
matrix has no zero rows.
  T
1 It is clear that if βi=0,
 
   T then the forced mode
A  WV ;  VB 2 has no contribution
   along the corresponding
 T eigenvector direction!
  n 
If βi≠0, then there always exists a choice of u(t) to define
the required contribution along the corresponding
eigenvector direction!
Slides by Anthony Rossiter
 11

NUMERICAL EXAMPLES

Slides by Anthony Rossiter

 12

Example 1
Is the following system controllable?
1 2   2
x  Ax  Bu; A    ; B 
1 0 3
First do the eigenvalue/vector decomposition.
I  A  0    2,1
  1 2   a  0   2
( A  2 I ) w1         w1   
 1  2   b  0  1 
 2 2   a  0   1
( A  I ) w2         w2   
1 1   b  0  1
Slides by Anthony Rossiter
 13

Example 1
The eigenvector matrix is therefore given as?
1 1
 
2  1 1  1 2
W   ; V W 
1 1  3
Therefore, one can find VB as:
1 1
Clearly all rows non-
 1 
2  2  0.33
VB   zero so controllable.
 3   2.67
3    
Slides by Anthony Rossiter
 14

Example 2
 2 1 3 1 
x  Ax  Bu ; A  3 1 4; B  2
1 0 1 0
First do the eigenvalue/vector decomposition.
I  A  0    4.24,0.24,0
  0 .6  Not paper and pen
( A  1 I ) w1 
 w1   0.78  computations so we
revert to MATLAB to
 0.18 demonstrate more.

Slides by Anthony Rossiter

 15

Example 2 continued
MATLAB is used to
solve:

A  WV ;
  T
1
 
 T
   VB
2
All non zero
   so
 T controllable.
  n 

Slides by Anthony Rossiter

 16

Example 3
 1.8 0.6  0.2 1
A   0.8 1.6  0.2; B  1
 0.4  0.8 2.6  0

Using MATLAB to
compute the
decomposition and VB
matrix gives: A zero row
so not fully
controllable.

Slides by Anthony Rossiter

 17

Summary
Used eigenvalue/vector decompositions to
introduce concepts of controllability/reachability
which means the ability to reach any desired value
of x(t) by judicious choice of u(t).

x  Ax  Bu A  WV
• Shown that full controllability requires the matrix
VB to have no zero rows.
• Not discussed non-simple Jordan forms/systems
with repeated eigenvalues.
Slides by Anthony Rossiter
 Anthony Rossiter
Department of Automatic Control and
Systems Engineering
University of Sheffield
www.shef.ac.uk/acse

© 2016 University of Sheffield

This work is licensed under the Creative Commons Attribution 2.0 UK: England & Wales Licence. To view a copy of this licence, visit
http://creativecommons.org/licenses/by/2.0/uk/ or send a letter to: Creative Commons, 171 Second Street, Suite 300, San Francisco, California 94105, USA.

It should be noted that some of the materials contained within this resource are subject to third party rights and any copyright notices must remain with these materials
in the event of reuse or repurposing.

If there are third party images within the resource please do not remove or alter any of the copyright notices or website details shown below the image.

(Please list details of the third party rights contained within this work.

If you include your institutions logo on the cover please include reference to the fact that it is a trade mark and all copyright in that image is reserved.)