Control Engineering

Ramkrishna Pasumarthy

Department of Electrical Engineering,

Indian Institute of Technology Madras
 Module 12
Lecture 4

Controllable Decomposition

Control Engineering Module 12 Lecture 4 Ramkrishna P. 1/10

Invariance with respect to
Similarity Transformations
Invariance with respect to Similarity Transformations

Consider the LTI systems

ẋ = Ax + Bu
, x ∈ Rn , u ∈ Rm (1)
x = Ax + Bu,
+

and a similarity transformation z = T−1 x, leading to

ż = Āz + B̄u
, Ā := T−1 AT, B̄ := T−1 B (2)
z = Āz + B̄u
+

Control Engineering Module 12 Lecture 4 Ramkrishna P. 2/10

Invariance with respect to Similarity Transformations

Theorem 7.5.1

The pair (A, B) is controllable if and only if the pair (Ā, B̄) = (T−1 AT, T−1 B) is control-
lable.

Proof Sketch

1. Express the controllability matrix of the transformed system in terms of the

controllability matrix of the original system.
2. Compare the rank of the controllability matrices.

Control Engineering Module 12 Lecture 4 Ramkrishna P. 3/10

Controllable Decomposition
Controllable Decomposition

Suppose the systems given in (1) is not completely state controllable. Then the rank of
the controllability matrix is less than n i.e.
( )
rank C = q < n

∴ ∃ q linearly independent column vectors in the controllability matrix.

Since, C is A-invariant ∃ a n × q matrix V whose columns constitute the basis vectors of C
and1
AV = VAc
[ ]
−1 Ac A12 [ ]
T AT = , where T = Vn×q Un×(n−q)
0 Au

and the columns of U are linearly independent of each other as well as linearly
independent of the columns of V.
1 refer Module 3 Lecture 2

Control Engineering Module 12 Lecture 4 Ramkrishna P. 4/10

Controllable Decomposition

{ }
Moreover, since Im B ⊂ C the columns of B can be written as a linear combination of
the columns of V.
∴ ∃ an q × m matrix Bc such that

B = VBc
[ ]
Bc
=⇒ B = T
0
[ ]
−1 Bc
=⇒ T B =
0

Control Engineering Module 12 Lecture 4 Ramkrishna P. 5/10

Controllable Decomposition
Theorem 7.5.2

For every LTI system (1), there is a similarity transformation that takes the system to
the form [ ] [ ]
Ac A12 Bc
T−1 AT = , T−1 B = , (3)
0 Au 0

for which
1. the controllable subspace of the transformed system (3) is given by
[ ]
Iq×q
C = Im
¯
0

and
2. the pair (Ac , Bc ) is controllable.

Control Engineering Module 12 Lecture 4 Ramkrishna P. 6/10

Example: RC Circuit

Consider the parallel RC circuit given in figure 1.

R1 R2
u

+ +
x1 C1 x2 C2
− −

Figure 1: Parallel RC cirucit

Control Engineering Module 12 Lecture 4 Ramkrishna P. 7/10

Example: RC Circuit

The state space representation of the RC circuit is given by

[ ] [ ]
− R11C1 0 1
ẋ = x + R11C1 u
0 − R21C2 R2 C2
[ ]
1 1 ω −ω 2
For R1 C1 = R2 C2 = ω, the controllability matrix C = . Therefore, rank(C) = 1.
ω −ω 2
[ ]T [ ]T
The basis of C is V = 1 1 . A vector U linearly independent to V and ∈ / C can be 1 − 1 .
[ ]
[ ] 1 1
Let T = V U = .
1 −1
[ ] [ ] [ ]
x1 + x2 −1 −ω 0 −1 ω
∴ z= , T AT = , T B=
x1 − x2 0 −ω 0

Control Engineering Module 12 Lecture 4 Ramkrishna P. 8/10

Example: RC Circuit

[ ] [ ]
Say y = x2 i.e. C = 0 1 . Therefore CT = 1 −1

Then,
[ ] [ ]
−ω 0 ω
ż = z+ u
0 −ω 0
[ ]
y = 1 −1 z

Therefore,
−1 −1 ω
T(s) = Cc (sI − Ac ) Bc = 1 (s + ω) ω=
s+ω

Control Engineering Module 12 Lecture 4 Ramkrishna P. 9/10

Overview

Summary: Module 12 Lecture 4 Contents: Module 12 Lecture 5

▶ Controllable Decomposition ▶ Observability

Control Engineering Module 12 Lecture 4 Ramkrishna P. 10/10