16.01.2015 Views

Lecture 9: Controllability and Observability - Illinois Institute of ...

Lecture 9: Controllability and Observability - Illinois Institute of ...

Lecture 9: Controllability and Observability - Illinois Institute of ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

Modern Control Systems<br />

Matthew M. Peet<br />

<strong>Illinois</strong> <strong>Institute</strong> <strong>of</strong> Technology<br />

<strong>Lecture</strong> 9: <strong>Controllability</strong> <strong>and</strong> <strong>Observability</strong>


<strong>Controllability</strong><br />

We would like to prove that<br />

To do this, we will prove that<br />

• Im(W t ) ⊂ R t<br />

• R t ⊂ Im(C(A, B))<br />

• Im(C(A, B)) ⊂ Im(W t )<br />

So far, we have show that<br />

R t = Im(W t ) = Im(C(A, B))<br />

Next, we will show that<br />

R t ⊂ C AB<br />

Im(W t ) ⊂ R t<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 2 / 15


This pro<strong>of</strong> M. Peet is especially useful, as it gives the<strong>Lecture</strong> control 9: <strong>Controllability</strong> input which will get us to x3 / 15<br />

<strong>Controllability</strong>: Im(W t ) ⊂ R t<br />

Theorem 1.<br />

Im(W t ) ⊂ R t<br />

Pro<strong>of</strong>.<br />

First, suppose that x ∈ Im(W t ) for some t > 0. Then x = W t z for some z.<br />

• Now let u(s) = B T e AT (t−s) z. Then<br />

Γ t u =<br />

=<br />

∫ t<br />

0<br />

∫ t<br />

= W t z = x<br />

• Thus x ∈ Im(Γ t ) = R t .<br />

We conclude that Im(W t ) ⊂ R t<br />

0<br />

e A(t−s) Bu(s)ds<br />

e A(t−s) BB T e AT (t−s) zds


Pro<strong>of</strong> by Contradiction: Im(C(A, B)) ⊂ Im(W t )<br />

The last pro<strong>of</strong> is a pro<strong>of</strong> by contradiction.<br />

Once you eliminate the impossible, whatever remains, no matter how<br />

improbable, must be the truth.<br />

There are two mutually exclusive possibilities<br />

1. x ∈ Im(C(A, B)) ⊂ Im(W t )<br />

2. There exists some x ∈ Im(C(A, B)) such that x ∉ Im(W t ) .<br />

We eliminate the second possibility by showing that:<br />

• If x ∉ Im(W t ) then x ∉ Im(C(A, B)).<br />

In shorth<strong>and</strong>:<br />

(¬2 ⇒ ¬1) ⇔ (1 → 2)<br />

An alternative would be to find an x which disproves the second possibility.<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 4 / 15


Pro<strong>of</strong> by Contradiction: Im(C(A, B)) ⊂ Im(W t )<br />

Theorem 2.<br />

Im(C(A, B)) ⊂ Im(W t ).<br />

Pro<strong>of</strong>.<br />

Suppose x ∉ Im(W t ). Then x ∈ Im(W t ) ⊥ .<br />

• As we have shown, this means x ∈ ker(W t ), so W t x = 0.<br />

• Thus<br />

x T W t x =<br />

=<br />

∫ t<br />

0<br />

∫ t<br />

0<br />

x T e A(t−s) BB T e AT (t−s) xds<br />

u(s) T u(s)ds = 0<br />

where u(s) = B T e AT (t−s) x.<br />

• This implies u(s) = B T e AT (t−s) x = 0 for all s ∈ [0, t].<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 5 / 15


Pro<strong>of</strong> by Contradiction: Im(C(A, B)) ⊂ Im(W t )<br />

Pro<strong>of</strong>.<br />

• This means that for all s ∈ [0, t],<br />

d k<br />

ds k BT e AT s x = B T ( A T ) k<br />

e<br />

A T s x = 0<br />

• At s = 0, this implies B T ( A T ) k<br />

x = 0 for all k.<br />

• We conclude that<br />

x T [ B AB · · · A n−1 B ] = 0<br />

• Thus C(A, B) T x = 0, so x ∈ ker C(A, B) T . As before, this means<br />

x ∈ Im(C(A, B)) ⊥ .<br />

• We conclude that x ∉ Im(C(A, B)). This proves by contradiction that<br />

Im(C(A, B)) ⊂ Im(W t ).<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 6 / 15


Summary: R t = Im(W t ) = Im(C(A, B))<br />

We have shown that<br />

R t = Im(W t ) = Im(C(A, B))<br />

Moreover, we have shown that for any x d ∈ R Tf , we can find a controller<br />

• Choose any z such that x d = W Tf z. (z = W −1<br />

T f<br />

x if W Tf is invertible)<br />

• Let u(t) = B T e AT (T f −t) z.<br />

• Then the system ẋ(t) = Ax(t) + B(t) with x(0) = 0 has solution with<br />

x(T f ) = x d .<br />

• x d = Γ Tf u.<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 7 / 15


Representation <strong>and</strong> <strong>Controllability</strong><br />

Question: Is the representation (A, B, C, D) <strong>of</strong> the system y = Gu,<br />

unique<br />

ẋ(t) = Ax(t) + Bu(t)<br />

y(t) = Cx(t) + Du(t)<br />

Question: Do there exist (Â, ˆB, Ĉ, ˆD) such that y <strong>and</strong> u also satisfy,<br />

ẋ(t) = Âx(t) + ˆBu(t)<br />

y(t) = Ĉx(t) + ˆDu(t)<br />

Answer: Of Course! Recall the similarity transform: z(t) = T x(t) for any<br />

invertible T . Then y <strong>and</strong> u also satisfy,<br />

ż(t) = T ẋ(t) = T Ax(t) + T Bu(t)<br />

= T AT −1 z(t) + T Bu(t)<br />

y(t) = Cx(t) + Du(t)<br />

= CT −1 x(t) + Du(t)<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 8 / 15


Representation <strong>and</strong> <strong>Controllability</strong><br />

Thus the pair (T AT −1 , T B, CT −1 , D) is also a representation <strong>of</strong> the map<br />

y = Gu.<br />

• Furthermore x(t) → 0 if <strong>and</strong> only if z(t) → 0.<br />

• So internal stability is unaffected.<br />

<strong>Controllability</strong> is Unaffected:<br />

C(T AT −1 , T B) = [ T B T AT −1 T B T AT −1 T AT −1 T B · · · T A n−1 B ]<br />

= T C(A, B)<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 9 / 15


Invariant Subspaces<br />

Definition 3.<br />

A subspace, W ⊂ X, is Invariant under the operator A : X → X if x ∈ W<br />

implies Ax ∈ W .<br />

For a linear operator, only subspaces can be invariant.<br />

Proposition 1.<br />

If W if A-invariant, then there exists an invertible T , such that<br />

]<br />

[ [Ā11<br />

T AT −1 Ā<br />

=<br />

12<br />

I<br />

<strong>and</strong> T W = Im<br />

Ā 21 Ā 22 0]<br />

That is, for any x ∈ W , T x =<br />

] [¯x1<br />

, which is clearly<br />

0<br />

Ā-invariant.<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 10 / 15


Invariant Subspaces<br />

Proposition 2.<br />

C AB is A-invariant.<br />

Pro<strong>of</strong>.<br />

The pro<strong>of</strong> is direct<br />

A · C(A, B) = A [ B AB · · · A n−1 B ] = [ AB A 2 B · · · A n B ]<br />

But, by Cayley-Hamilton,<br />

A n =<br />

n−1<br />

∑<br />

a i A i<br />

so if x ∈ C AB , there exists a z such that<br />

Ax = A [ B AB · · · A n−1 B ] z<br />

⎡<br />

⎤<br />

z n a 0<br />

= [ B · · · A n−1 B ] z 1 + z n a 1<br />

⎢<br />

⎥<br />

⎣ . ⎦ ∈ C AB<br />

z n+1 + z n a n−1<br />

i=0<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 11 / 15


<strong>Controllability</strong> Form<br />

Since C AB is an invariant subspace <strong>of</strong> A, there exists an invertible T such that<br />

]<br />

[Ā11<br />

T AT −1 Ā<br />

=<br />

12<br />

0 Ā 22<br />

[¯x<br />

<strong>and</strong> T x = for any x ∈ C<br />

0]<br />

AB .<br />

• Clearly B ∈ C AB .<br />

[ ] ¯B1<br />

• Thus T B = .<br />

0<br />

Definition 4.<br />

The pair (A, B) is in <strong>Controllability</strong> Form when<br />

[ ]<br />

A11 A<br />

A =<br />

12<br />

<strong>and</strong> B =<br />

0 A 22<br />

<strong>and</strong> the pair (A 11 , B 1 ) is controllable.<br />

[ ]<br />

B1<br />

,<br />

0<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 12 / 15


<strong>Controllability</strong> Form<br />

When a system is in controllability form, the dynamics have special structure<br />

ẋ 1 (t) = A 11 x 1 (t) + A 12 x 2 (t) + B 1 u(t)<br />

ẋ 2 (t) = A 22 x 2 (t)<br />

The uncontrolled dynamics are autonomous.<br />

• Cannot be stabilized or controlled.<br />

We can formulate a procedure for putting a system in <strong>Controllability</strong> Form<br />

1. Find an orthonormal basis, [ ]<br />

v 1 · · · v r for CAB .<br />

2. Complete the basis in R n : [ ]<br />

v r+1 · · · v n .<br />

3. Define T = [ ]<br />

v 1 · · · v n .<br />

4. Construct Ā = T AT −1 <strong>and</strong> ¯B = T B<br />

◮ Works for ANY invariant subspace.<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 13 / 15


<strong>Controllability</strong> Form<br />

Example<br />

Let<br />

⎡<br />

1 −1<br />

⎤<br />

0<br />

⎡ ⎤<br />

1<br />

A = ⎣−1 1 0 ⎦ <strong>and</strong> B = ⎣0⎦<br />

0 0 −1<br />

0<br />

Construct C(A, B) = [ B AB A 2 B ] .<br />

⎡<br />

1 −1<br />

⎤ ⎡ ⎤<br />

0 1<br />

⎡ ⎤<br />

1<br />

AB = ⎣−1 1 0 ⎦ ⎣0⎦ = ⎣−1⎦ ,<br />

0 0 −1 0 0<br />

⎡<br />

⎤ ⎡ ⎤ ⎡ ⎤<br />

1 −1 0 1 0<br />

A 2 B = A(AB) = ⎣−1 1 0 ⎦ ⎣−1⎦ = ⎣0⎦<br />

0 0 −1 0 0<br />

Thus<br />

⎡<br />

1 1<br />

⎤<br />

0<br />

C(A, B) = ⎣0 −1 0⎦<br />

0 0 0<br />

Thus rank C(A, B) = 2 < n = 3 which means not controllable.<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 14 / 15


<strong>Controllability</strong> Form<br />

Example Continued<br />

Using Gramm-Schmidt, we can construct an orthonormal basis for C AB<br />

⎧⎡<br />

⎤ ⎡ ⎤⎫<br />

⎧⎡<br />

⎤ ⎡ ⎤⎫<br />

⎨ 1 1 ⎬ ⎨ 1 0 ⎬<br />

C AB = span ⎣0⎦ , ⎣−1⎦<br />

⎩<br />

⎭ = span ⎣0⎦ , ⎣1⎦<br />

⎩ ⎭ = span {v 1, v 2 }<br />

0 0<br />

0 0<br />

Let v 3 = [ 0 0 1 ] T<br />

. Then<br />

⎡<br />

T −1 = ⎣ 1 0 0<br />

⎤<br />

0 1 0⎦ = I<br />

0 0 1<br />

So T = I, which is because the system is already in controllability form. We<br />

could also have used<br />

⎡<br />

0 1<br />

⎤<br />

0<br />

⎡<br />

1 −1<br />

⎤<br />

0<br />

T −1 = ⎣1 0 0⎦ to get T AT −1 = ⎣−1 1 0 ⎦ = A<br />

0 0 1<br />

0 0 −1<br />

M. Peet <strong>Lecture</strong> 9: <strong>Controllability</strong> 15 / 15

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!