State Space Methods

Lecture 1: State space models

Jakob Stoustrup
jakob@control.aau.dk www.control.aau.dk/jakob/

Automation & Control Department of Electronic Systems Aalborg University Denmark

Contents
One slide course summary State space models Example: mass-spring-damper State space models and transfer functions Poles and zeros of state space models State space transformations

One slide course summary

State space models

One slide course summary

State space models Controllability

One slide course summary

State space models Controllability State feedback design (pole assignment)

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment)

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment) Observer based control (separation theorem)

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment) Observer based control (separation theorem) Reduced order observers

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment) Observer based control (separation theorem) Reduced order observers Integral state space control

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment) Observer based control (separation theorem) Reduced order observers Integral state space control Zero assignment

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment) Observer based control (separation theorem) Reduced order observers Integral state space control Zero assignment Anti-windup

One slide course summary

State space models Controllability State feedback design (pole assignment) Observability Observer gain design (pole assignment) Observer based control (separation theorem) Reduced order observers Integral state space control Zero assignment Anti-windup Optimal control

Contents
One slide course summary State space models Example: mass-spring-damper State space models and transfer functions Poles and zeros of state space models State space transformations

State space models

A linear third order system in continuous time with two inputs and two outputs has a state space model of the following form: x1 = a11 x1 + a12 x2 + a13 x3 + b11 u1 + b12 u2 x2 = a21 x1 + a22 x2 + a23 x3 + b21 u1 + b22 u2 x3 = a31 x1 + a32 x2 + a33 x3 + b31 u1 + b32 u2 where x1 , x2 , x3 are called the states, u1 , u2 are called the inputs, and y1 , y2 are called the outputs.

State space models

A linear third order system in continuous time with two inputs and two outputs has a state space model of the following form: x1 = a11 x1 + a12 x2 + a13 x3 + b11 u1 + b12 u2 x2 = a21 x1 + a22 x2 + a23 x3 + b21 u1 + b22 u2 x3 = a31 x1 + a32 x2 + a33 x3 + b31 u1 + b32 u2 System equations

State space models

A linear third order system in continuous time with two inputs and two outputs has a state space model of the following form: x1 = a11 x1 + a12 x2 + a13 x3 + b11 u1 + b12 u2 x2 = a21 x1 + a22 x2 + a23 x3 + b21 u1 + b22 u2 x3 = a31 x1 + a32 x2 + a33 x3 + b31 u1 + b32 u2 y1 = c11 x1 + c12 x2 + c13 x3 + d11 u1 + d12 u2 y2 = c21 x1 + c22 x2 + c23 x3 + d21 u1 + d22 u2

State space models

A linear third order system in continuous time with two inputs and two outputs has a state space model of the following form: x1 = a11 x1 + a12 x2 + a13 x3 + b11 u1 + b12 u2 x2 = a21 x1 + a22 x2 + a23 x3 + b21 u1 + b22 u2 x3 = a31 x1 + a32 x2 + a33 x3 + b31 u1 + b32 u2 y1 = c11 x1 + c12 x2 + c13 x3 + d11 u1 + d12 u2 y2 = c21 x1 + c22 x2 + c23 x3 + d21 u1 + d22 u2 Output equations

State space models

In matrix form, a continuous time state space model can be written as: x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) where: x1 x = x 2 x3 , u= u1 u2 , y= y1 y2

State space models

In matrix form, a continuous time state space model can be written as: x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) where: a11 a12 a13 A = a21 a22 a23 a31 a32 a33 , b11 b12 B = b21 b22 b31 b32

State space models

In matrix form, a continuous time state space model can be written as: x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) where: C= c11 c12 c13 c21 c22 c23 , D= d11 d12 d21 d22

State space models

In matrix form, a continuous time state space model can be written as: x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) Similarly, a discrete time state space model can be written as: x(k + 1) = Ax(k) + Bu(k) y(k) = Cx(k) + Du(k)

Choosing state variables

For a physical system, the number of states required is typically equal to the number of energy storages, and a possible choice of state variables is often those variables, that represent energy storage.

Choosing state variables

Linear component Recommended variable condensator electrical coil spring mass (kinetic) mass (potential) inertia wheel plane spring heat storage gas accumulator voltage current length velocity elevation angular velocity winding angle temperature pressure

Contents
One slide course summary State space models Example: mass-spring-damper State space models and transfer functions Poles and zeros of state space models State space transformations

Example: mass-spring-damper
p k m F

friction coef.: c The force F is considered as input, and the mass velocity v is considered as output of this system. The system is of second order, since it has one mass which can contain both kinetic and potential energy.

Example: mass-spring-damper
A possible selection of states are the position p and the velocity v.

Example: mass-spring-damper
A possible selection of states are the position p and the velocity v. The derivative of v is given by Newtons second law: mv = k p c v + F = c 1 k v = p v+ F m m m

Example: mass-spring-damper
A possible selection of states are the position p and the velocity v. The derivative of v is given by Newtons second law: mv = k p c v + F = c 1 k v = p v+ F m m m The derivative of p is simply given by: p=v

Example: mass-spring-damper
Thus, we have the following state space model: p v v = = 0 k m 0 1 c m 1 p v p v + + 0
1 m

F F

which is indeed of the form: x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t)

Contents
One slide course summary State space models Example: mass-spring-damper State space models and transfer functions Poles and zeros of state space models State space transformations

State space model transfer fct.

Taking Laplace transforms of the system x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) yields

State space model transfer fct.

Taking Laplace transforms of the system x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) yields sx(s) = Ax(s) + Bu(s) y(s) = Cx(s) + Du(s) rearranging, we obtain:

State space model transfer fct.

Taking Laplace transforms of the system x(t) = Ax(t) + Bu(t) y(t) = Cx(t) + Du(t) yields sx(s) = Ax(s) + Bu(s) y(s) = Cx(s) + Du(s) rearranging, we obtain: (sI A) x(s) = Bu(s) y(s) = Cx(s) + Du(s)

State space model transfer fct.

sx(s) = Ax(s) + Bu(s) y(s) = Cx(s) + Du(s) rearranging, we obtain: (sI A) x(s) = Bu(s) y(s) = Cx(s) + Du(s) Premultiplying with (sI A) on either side of the system equation, results in x(s) = (sI A)1 Bu(s) y(s) = Cx(s) + Du(s)
1

State space model transfer fct.

(sI A) x(s) = Bu(s) y(s) = Cx(s) + Du(s) Premultiplying with (sI A)1 on either side of the system equation, results in x(s) = (sI A)1 Bu(s) y(s) = Cx(s) + Du(s) Finally, we obtain: x(s) = (sI A) Bu(s) y(s) = C (sI A)1 Bu(s) + Du(s)
1

State space model transfer fct.

Premultiplying with (sI A)1 on either side of the system equation, results in x(s) = (sI A)1 Bu(s) y(s) = Cx(s) + Du(s) Finally, we obtain: x(s) = (sI A)1 Bu(s) y(s) = C (sI A)1 Bu(s) + Du(s) Consequently, y(s) = G(s)u(s) , where: G(s) = C (sI A)
1

B+D

Example: mass-spring-damper
For the spring-mass-damper system with m = 1, c = 3, k = 2, the state space representation is: p v v = = 0 2 0 1 3 1 p v p v + + 0 1 0 F F

Thus, the transfer function becomes: G(s) = C (sI A)1 B + D

Example: mass-spring-damper
G(s) = C (sI A) = 0 1 s 0 0 s
1

B+D 0 1 2 3
1

0 1

+ 0

Example: mass-spring-damper
G(s) = C (sI A) = 0 1 s 0 0 s s 1 2 s+3
1

B+D 0 1 2 3 0 1
1

0 1

+ 0

= 0 1

Example: mass-spring-damper
G(s) = C (sI A) = 0 1 s 0 0 s s 1 2 s+3
1

B+D 0 1 2 3 0 1 s+3 1 2 s 0 1
1

0 1

+ 0

= 0 1

1 = 2 0 1 s + 3s + 2

Example: mass-spring-damper
G(s) = C (sI A) = 0 1 s 0 0 s s 1 2 s+3
1

B+D 0 1 2 3 0 1 s+3 1 2 s 0 1
1

0 1

+ 0

= 0 1

1 = 2 0 1 s + 3s + 2 s = 2 s + 3s + 2

Transfer fct. state space model

Consider the transfer function g(s) = From the relationship 1 y(s) = 2 u(s) s + a1 s + a2 we infer s2 y(s) + a1 sy(s) + a2 y(s) = u(s) Taking inverse Laplace transform, this becomes: y (t) + a1 y(t) + a2 y(t) = u(t)
1 s2 +a1 s+a2 .

Transfer fct. state space model

y (t) + a1 y(t) + a2 y(t) = u(t) A possible choice of states is: x1 = y, x2 = y. With this choice, the system equations become: x1 = y = x2 x2 = y = a1 y a2 y + u = a2 x1 a1 x2 + u In matrix form, we obtain:

Transfer fct. state space model

y (t) + a1 y(t) + a2 y(t) = u(t) A possible choice of states is: x1 = y, x2 = y. With this choice, the system equations become: x1 = y = x2 x2 = y = a1 y a2 y + u = a2 x1 a1 x2 + u In matrix form, we obtain: x1 0 1 = a2 a1 x2 y = 1 0 x1 x2 x1 x2 0 u 1 0 u

+ +

Contents
One slide course summary State space models Example: mass-spring-damper State space models and transfer functions Poles and zeros of state space models State space transformations

Poles of state space models

With G(s) = C (sI A)1 B + D we have that: G(s) for s p Hence, p is a pole for G(s) det (pI A) = 0

Poles of state space models

With G(s) = C (sI A)1 B + D we have that: G(s) for s p Hence, p is a pole for G(s) p is an eigenvalue for A det (pI A) = 0

Example: mass-spring-damper
For the mass-spring-damper system, the A matrix was: A= 0 2 1 3

which has the characteristic polynomial: 1 det (I A) = 2 +3 = 2 + 3 + 2 = ( + 1)( + 2) Thus, the system has poles in {1, 2}.

Zeros of state space models

With G(s) = C (sI A)1 B + D we have that: G(z)u = 0 C (zI A)
1

Bu + Du = 0
1

C + Du = 0 , = (zI A) A zI B C D u

Bu

C + Du = 0 , (A zI) + Bu = 0 =0

Zeros of state space models

G(z)u = 0 C (zI A)
1

Bu + Du = 0
1

C + Du = 0 , = (zI A) A zI B C D u

Bu

C + Du = 0 , (A zI) + Bu = 0 =0

Thus, z is a zero for G(s) A zI B C D does not have full column rank

Example: mass-spring-damper
For the mass-spring-damper system, zeros must satisfy: A zI B =0 C D or 0 z 1 z 1 (1) = z = 0 2 3 z 1 = 0 1 0 1 0 Hence, the system has a zero in the origin.

Contents
One slide course summary State space models Example: mass-spring-damper State space models and transfer functions Poles and zeros of state space models State space transformations

State space transformations

State space representations are not unique! Given one model: x = Ax + Bu y = Cx + Du another model can be obtained by a non-singular transformation of the state vector: x = T , = T 1 x

State space transformations

Introducing this in the state space model, we obtain: T = AT + Bu y = CT + Du or, equivalently = T 1 AT + T 1 Bu y = CT + Du

State space transformations

= T 1 AT + T 1 Bu y = CT + Du Thus, a new state space model of the form = A + Bu y = C + Du where A = T 1 AT B = T 1 B C = CT D=D has been obtained.

Example: mass-spring-damper
For the mass-spring-damper system, we change basis using the following transformation matrix: T = 1 1 1 2 , T
1

2 1 1 1

This gives the following new state space representation: = A + Bu y = C + Du with

Example: mass-spring-damper
A = T 1 AT = 2 1 1 1 0 1 2 3 1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = 2 1 1 1 1 0 0 2 0 1 2 3 1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = B = T 1 B = 2 1 1 1 1 0 0 2 2 1 1 1 0 1 0 1 2 3 1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = B = T 1 B = 2 1 1 1 1 0 0 2 2 1 1 1 0 1 = 1 1 0 1 2 3 1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = B = T 1 B = 2 1 1 1 1 0 0 2 2 1 1 1 0 1 = 1 1 0 1 2 3 1 1 1 2

C = CT = 0 1

1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = B = T 1 B = 2 1 1 1 1 0 0 2 2 1 1 1 0 1 = 1 1 = 1 2 0 1 2 3 1 1 1 2

C = CT = 0 1

1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = B = T 1 B = 2 1 1 1 1 0 0 2 2 1 1 1 0 1 = 1 1 = 1 2 0 1 2 3 1 1 1 2

C = CT = 0 1 D=D

1 1 1 2

Example: mass-spring-damper
A = T 1 AT = = B = T 1 B = 2 1 1 1 1 0 0 2 2 1 1 1 0 1 = 1 1 = 1 2 0 1 2 3 1 1 1 2

C = CT = 0 1 D=D=0

1 1 1 2

Example: mass-spring-damper
Transfer matrix: G(s) = C sI A
1

B+D

Example: mass-spring-damper
Transfer matrix: G(s) = C sI A = 1 2
1

B+D
1

s+1 0 0 s+2

1 1

+0

Example: mass-spring-damper
Transfer matrix: G(s) = C sI A = 1 2
1

B+D
1

s+1 0 0 s+2

1 1

+0

1 1 (s + 2) + 2(s + 1) = +2 = s+1 s+2 (s + 1)(s + 2)

Example: mass-spring-damper
Transfer matrix: G(s) = C sI A = 1 2
1

B+D
1

s+1 0 0 s+2

1 1

+0

1 1 (s + 2) + 2(s + 1) = +2 = s+1 s+2 (s + 1)(s + 2) s = 2 s + 3s + 2

Example: mass-spring-damper
Transfer matrix: G(s) = C sI A = 1 2
1

B+D
1

s+1 0 0 s+2

1 1

+0

1 1 (s + 2) + 2(s + 1) = +2 = s+1 s+2 (s + 1)(s + 2) s = 2 s + 3s + 2 = G(s)