Académique Documents
Professionnel Documents
Culture Documents
representations
(SEM)
representations Modelling of
dynamical
systems
Properties
Discrete-time
O. Sename1 systems
State feedback
1 Gipsalab, control
CNRS-INPG, FRANCE
Olivier.Sename@inpg.fr Observer
www.lag.ensieg.inpg.fr/sename Integral Control
A polynomial
approach
Approche Etat pour la commande / IEG- SEM
Further in
discrete-time
control
Conclusion
State space
Outline representations
(SEM)
O.Sename
Introduction
Introduction
Discrete-time
Discrete-time systems systems
State feedback
State feedback control control
Observer
Observer Integral Control
A polynomial
Integral Control approach
Further in
A polynomial approach discrete-time
control
Conclusion
Further in discrete-time control
Conclusion
State space
References representations
(SEM)
O.Sename
Introduction
Some interesting books: Modelling of
dynamical
◮ K.J. Astrom and B. Wittenmark, Computer-Controlled systems
O.Sename
Nominal stability (NS): The system is stable with the
Introduction
nominal model (no model uncertainty)
Modelling of
Nominal Performance (NP): The system satisfies the dynamical
systems
performance specifications with the nominal Properties
model (no model uncertainty) Discrete-time
systems
Robust stability (RS): The system is stable for all State feedback
perturbed plants about the nominal model, up control
Integral Control
(including the real plant)
A polynomial
Robust performance (RP): The system satisfies the approach
real plant).
State space
representations
(SEM)
O.Sename
Introduction
Recall of the "control design" process: Modelling of
dynamical
◮ Plant study and modelling systems
Properties
◮ Determination of sensors and actuators (measured
Discrete-time
and controlled outputs, control inputs) systems
Integral Control
◮ Simulation tests
A polynomial
◮ Implementation, tests and validation approach
Further in
discrete-time
control
Conclusion
State space
representations
Different issues for modelling: (SEM)
A polynomial
◮ Formulate a nonlinear state-space model, i.e. a matrix approach
O.Sename
Why state space equations ?
Introduction
◮ dynamical systems where physical equations can be
Modelling of
derived : electrical engineering, mechanical dynamical
systems
engineering, aerospace engineering, microsystems,
Properties
process plants .... Discrete-time
systems
◮ include physical parameters: easy to use when
State feedback
parameters are changed for design control
Integral Control
◮ Easy to extend to Multi-Input Multi-Output (MIMO) A polynomial
systems approach
Further in
◮ Advanced control design method are based on state discrete-time
control
space equations (reliable numerical optimisation tools)
Conclusion
State space
Some physical examples representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
General dynamical system representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
A polynomial
where f and g are non linear functions. approach
Further in
discrete-time
control
Conclusion
State space
Example: Inverted pendulum representations
(SEM)
It is described by: O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Parameters:
Further in
discrete-time
control
Conclusion
State space
Example: Inverted pendulum representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
The dynamical equations are as follows: systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Example: Lateral vehicle model representations
(SEM)
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Definition of state space representations representations
(SEM)
O.Sename
Introduction
A continuous-time LINEAR state space system is given Modelling of
as : ( dynamical
systems
ẋ (t) = A(x(t) + Bu(t), x(0) = x0
Properties
(2)
y(t) = Cx(t) + Du(t) Discrete-time
systems
O.Sename
Modelling of
dynamical
di
Ri + L + e = u e = Ke ω systems
dt Properties
dω
= −f ω + Γm Γm = K c i
Discrete-time
J systems
dt State feedback
control
System of 2 equations
of order
1 =⇒ 2 state variables. A Observer
ω
possible choice x = It gives: Integral Control
i A polynomial
approach
Further in
−f /J Kc /J 0 discrete-time
A= B= C= 0 1 control
−Ke /L −R/L 1/L
Conclusion
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
A linearisation within 2 regions gives State space
representations
(SEM)
ẋ(t) = Ax(t) + Bu(t) + Ed (t) O.Sename
Modelling of
with dynamical
systems
γ −Cd Cd
α
−1 Properties
Irot Irot Irot 0 Irot
Discrete-time
A = Kd 0 −Kd , B = 0 , E = 0 , systems
Cd −1 Cd −1
Igen Igen Igen Igen 0 State feedback
control
Observer
and C = 0 0 1 Integral Control
x1 = rotor-speed x2 = drive-train torsion spring force, x3 = A polynomial
rotational generator speed approach
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
where x1 is the relative position, M1 the system mass, k1 discrete-time
control
the spring coefficient, u the force generated by the active
Conclusion
damper, and F1 is an external disturbance. Applying the
mechanical equations it leads:
M1 ẍ1 = −k1 x1 + u + F1 (3)
State space
Examples: Suspension cont. representations
(SEM)
O.Sename
Introduction
Modelling of
x1 dynamical
The choice x = gives systems
ẋ1
Properties
Discrete-time
ẋ(t) = Ax(t) + Bu(t) + Ed (t) systems
Observer
where d = F1 , y = x1 with
Integral Control
A polynomial
0 1 0 approach
A= , B=E = , and C = 0 1
−k1 /M1 0 1/M1 Further in
discrete-time
control
Conclusion
State space
Exercice representations
(SEM)
O.Sename
Introduction
Let the following quarter car model with active suspension. Modelling of
dynamical
Zcaisse and Zroue ) are the relative systems
position of the chassis and of the Properties
tire, Discrete-time
systems
mc (resp. mr ) the mass of the chas-
State feedback
sis (resp. of the tire), control
A polynomial
u the active damper force, approach
Zsol is the road profile. Further in
discrete-time
Choose some state variables and give a state space control
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
Linearisation systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Equilibrium point representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
State feedback
0 = f ((xeq (t), ueq (t), t) (4) control
Observer
For the pendulum, we can choose y = θ = f = 0. Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Linearisation Method (1) representations
(SEM)
O.Sename
The linearisation can be done around an equilibrium point
Introduction
or around a particular point.
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Linearisation Method (2) representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
This leads to a linear state space representation of the systems
A polynomial
with A = ∂∂xf |x=xeq ,u=ueq , B = ∂∂uf |x=xeq ,u=ueq , approach
Conclusion
State space
Example: Inverted pendulum (2) representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
Applying the linearisation method leads to : systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
State feedback
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Equivalence transfer function - state space representations
(SEM)
representation O.Sename
Introduction
Consider a linear system given by:
Modelling of
dynamical
ẋ(t) = Ax(t) + Bu(t), x(0) = x0 systems
(6)
y(t) = Cx(t) + Du(t) Properties
Discrete-time
systems
Using the Laplace transform (and assuming zero initial
State feedback
condition x0 = 0), (6) becomes: control
Observer
A polynomial
approach
Then the transfer function matrix of system (6) is given by
Further in
discrete-time
N(s) control
G(s) = C(sIn − A)−1 B + D = (7) Conclusion
D(s)
O.Sename
y b s2 + b s + b3 N(s) Observer
= G(s) = 3 1 s 2 = Integral Control
u s + a12 + a2 s + a3 D(s)
A polynomial
approach
Numerator equal to denominator order Further in
discrete-time
control
y b0 s3 + b1 s2 + b2 s + b3 N(s)
= G(s) = = Conclusion
u s3 + as12 + a2 s + a3 D(s)
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Canonical forms representations
(SEM)
O.Sename
Further in
c0 + c1 s + . . . + cn−1 sn−1 discrete-time
G(s) = control
a0 + a1 s + . . . + an−1 sn−1 + sn Conclusion
b1 b2 bn
G(s) = + + ... + Introduction
s − a1 s − a2 s − an Modelling of
dynamical
systems
◮ Define a set of transfer functions:
Properties
Xi (s) bi
= ⇒ ẋi = ai xi bi ui Discrete-time
U(s) s − ai systems
State feedback
◮ This gives control
Observer
ẋ(t) = Ax(t) + Bu(t), x(0) = x0 Integral Control
(8)
y(t) = Cx(t) + Du(t) A polynomial
approach
a1 0 . . . 0 b1 Further in
discrete-time
.. b2 control
0 a2 0 .
with A = , B =
.. and Conclusion
0 . .
.. .. 0 .
0 . . . 0 an bn
C= 1 1 1 1 .
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
State feedback
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Solution of state space equations - continuous representations
(SEM)
case O.Sename
Introduction
Modelling of
dynamical
The state x(t), solution of ẋ(t) = Ax(t), with initial systems
Properties
condition x(0) = x0 is given by
Discrete-time
systems
At
x(t) = e x(0) (9) State feedback
control
A polynomial
1. Inverse Laplace transform of (sIn − A)−1 : approach
Further in
2. Diagonalisation of A discrete-time
control
3. Cayley-Hamilton method Conclusion
State space
Complete state solution representations
(SEM)
O.Sename
Introduction
|0
| {z } Discrete-time
{z } systems
free response
forced response State feedback
control
Integral Control
Simulation of state space systems A polynomial
approach
Use lsim.
Further in
Example: discrete-time
control
t = 0:0.01:5; u = sin(t); lsim(sys,u,t)
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Non unicity representations
(SEM)
Given a transfer function, there exists an infinity of state O.Sename
Discrete-time
State feedback
consider the change of variables x = Tz (T being an control
A polynomial
T ż(t) = ATz(t) + Bu(t) (12) approach
Further in
y(t) = CTz(t) + Du(t) (13) discrete-time
control
Hence Conclusion
O.Sename
Modelling of
function of the previous system is: dynamical
systems
−1 −1 −1 Discrete-time
= C T (sIn − T AT ) T B +D (17) systems
State feedback
(18) control
Observer
Using In = T −1 T , we get Integral Control
A polynomial
G̃(s) = C T T −1 (sIn − A)−1 T T −1 B + D = G(s) (19) approach
Further in
discrete-time
control
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
Stability systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Stability representations
(SEM)
O.Sename
Definition Introduction
An equilibrium point xeq is stable if, for all ρ > 0, there Modelling of
dynamical
exists a η > 0 such that: systems
Properties
State feedback
control
Definition Observer
Further in
kx(0) − xeq k < η =⇒ x(t) → xeq , when t → ∞ discrete-time
control
Conclusion
These notions are equivalent for linear systems (not for
non linear ones).
State space
Stability Analysis representations
(SEM)
O.Sename
Introduction
The system poles are then the eigenvalues of the matrix A. Properties
stable if Re(λi ) < 0, ∀i, where λi , ∀i, are the eigenvalues of Integral Control
A. A polynomial
approach
Using Matlab, if SYS is an SS object then pole(SYS) Further in
discrete-time
computes the poles P of the LTI model SYS. It is control
equivalent to compute eig(A). Conclusion
State space
Stability Analysis - Lyapunov representations
(SEM)
Theorem Properties
Observer
T
A P + PA = −Q (20) Integral Control
A polynomial
approach
see lyap in MATLAB. Further in
Proof: The Lyapunov theory says that a linear system is discrete-time
control
stable if there exists a continuous function V (x) s.t.: Conclusion
dV
V (x) > 0 with V (0) = 0 and V̇ (x) = ≥0
dx
A possible Lyapunov function for the above system is :
State space
About zeros representations
(SEM)
O.Sename
Properties
them using a state space model.
Discrete-time
◮ Zero: is a generalized frequency α for which the systems
A polynomial
A − λ In B approach
=0 (21) Further in
C D discrete-time
control
Conclusion
In Matlab use zero
s+3
Example: find the zero of : s2+5s+2
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
Controllability systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Controllability representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Controllability refers to the ability of controlling a Properties
state-space model using state feedback. Discrete-time
systems
Definition State feedback
control
Given two states x0 and x1 , the system (6) is controllable if
Observer
there exist t1 > 0 and a piecewise-continuous control input
Integral Control
u(t), t ∈ [0, t1 ], such that x(t) takes the values x0 for t = 0 A polynomial
and x1 for t = t1 . approach
Further in
discrete-time
control
Conclusion
State space
Controllability cont. representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
Proposition systems
Integral Control
Using Matlab, if SYS is an SS object then crtb(SYS) A polynomial
approach
returns the controllability matrix of the state-space model
Further in
SYS with realization (A,B,C,D). This is equivalent to discrete-time
control
ctrb(sys.a,sys.b)
Conclusion
State space
Exercices representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
Test the controllability of the previous examples: DC motor,
State feedback
suspension, inverted pendulum. control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
Observability systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Observability representations
(SEM)
O.Sename
Introduction
Modelling of
Observability refers to the ability to estimate a state dynamical
systems
variable.
Properties
Definition Discrete-time
systems
A linear system (2) is completely observable if, given the State feedback
control and the output over the interval t0 ≤ t ≤ T , one can control
Integral Control
It is equivalent to characterize the non-observability as :
A polynomial
A state x(t) is not observable if the corresponding output approach
Conclusion
State space
Where does observability come from ? representations
(SEM)
O.Sename
Introduction
ẋ = −x + u Properties
Discrete-time
y = 2x systems
State feedback
control
and
Observer
−1 0 1 Integral Control
ẋ = x+ u
0 −2 1 A polynomial
approach
Further in
y = 2 0 x discrete-time
control
Conclusion
State space
Observability cont. representations
(SEM)
O.Sename
Introduction
Proposition Modelling of
dynamical
C systems
CA
Properties
The observability matrix is defined by O = .. . Discrete-time
. systems
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
Minimality systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Minimality representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Definition Properties
A state space representation of a linear system (2) of order Discrete-time
systems
n is said to be minimal if it is controllable and observable.
State feedback
In this case, the corresponding transfer function G(s) is of control
Observer
minimal order n, i.e is irreducible (no cancellation of poles
Integral Control
and zeros).
A polynomial
When the transfer function is not of minimal order, there approach
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Kalman decomposition representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
When the linear system Discrete-time
(2) is not completely systems
Conclusion
State space
Toward digital control representations
(SEM)
as: Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
This requires the use of the discrete theory. Further in
m (Sampling theory + Z-Transform) m discrete-time
control
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
Z-Transform systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Definitions representations
(SEM)
O.Sename
Mathematical definition
Because the output of the ideal sampler, x ∗ (t), is a series Introduction
Modelling of
of impulses with values x(kTe ), we have: dynamical
systems
∞
∑ x(kTe )δ (t − kTe )
Properties
x ∗ (t) = Discrete-time
k =0 systems
State feedback
by using the Laplace transform, control
Observer
∞
∑ x(kTe )e−ksT
Integral Control
L [x ∗ (t)] = e
A polynomial
k =0 approach
Further in
Noting z = esTe , we can derive the so called Z-Transform discrete-time
control
∞ Conclusion
k =0
State space
Properties representations
(SEM)
O.Sename
Definition Introduction
∞ Modelling of
X (z) = Z [x(k)] = ∑ x(k)z −k dynamical
systems
k =0 Properties
Discrete-time
Properties systems
State feedback
Z [α x(k) + β y(k)] = α X (z) + β Y (z) control
Observer
Z [x(k − n)] = z −n Z [x(k)] Integral Control
d A polynomial
Z [kx(k)] = −z Z [x(k)] approach
dz
Further in
Z [x(k) ∗ y(k)] = X (z).Y (z) discrete-time
control
lim x(k) = lim (z − 1)X (z) Conclusion
k →∞ 1→z −1
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Exercise representations
(SEM)
Determine the Z-Transform of the step function (1) and of O.Sename
the ramp function (2)
Introduction
Properties
Discrete-time
Solution systems
Observer
1 z
Xstep (z) = 1 + z −1 + z −2 + · · · = = Integral Control
1 − z −1 z − 1 A polynomial
approach
2) Ramp (note that xramp (k) = kxstep (k)) Further in
discrete-time
control
d z
Xramp (z) = −z Conclusion
dz z − 1
z
=
(z − 1)2
State space
Zero order holder representations
(SEM)
O.Sename
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Zero order holder (cont’d) representations
(SEM)
O.Sename
Introduction
Properties
1 e−sTe Discrete-time
GBOZ (s) = − systems
s s
State feedback
1−e e−sT control
=
s Observer
Integral Control
A polynomial
approach
Influence of the D/A and A/D
Further in
Note that the precision is also limited by the available discrete-time
control
precision of the converters (either A/D or D/A).
Conclusion
This error is also called the amplitude quantization error.
State space
Representation of the discrete linear systems representations
(SEM)
∞ Introduction
y(k) = ∑ h(k − n)u(n) Modelling of
dynamical
n=0
systems
Discrete-time
systems
Y (z) = Z [h(k)]U(z) = H(z)U(z) State feedback
control
Observer
b0 + b1 z + · · · + bm z m Y
H(z) = n
= Integral Control
a0 + a1 z + · · · + an z U A polynomial
approach
where n (≥ m) is the order of the system Further in
discrete-time
Corresponding difference equation: control
Conclusion
1
y(k) = b0 u(k − n) + b1u(k − n + 1) + · · · + bm u(k − n + m)
an
− a0 y(k − n) − a2y(k − n + 1) − · · · − an−1y(k − 1)
State space
Some useful transformations representations
(SEM)
O.Sename
x (t) X (s) X (z)
δ (t) 1 1 Introduction
z(1−e−aTe )
1 − e−at 1
s(s+a) (z−1)(z−e−aTe )
State feedback
control
ω zsin(ω Te )
sin(ω t) s 2 +ω 2 z 2 −2zcos(ω Te )+1
Observer
Further in
Exercise discrete-time
control
Discretize (sampling time Te ) the system described by the
Conclusion
Laplace function (using a Zero order holder):
Y (s) 1
H(s) = =
U(s) s(s + 1)
State space
Exercise representations
(SEM)
Discretize the system described by the Laplace function O.Sename
(using a Zero order holder):
Introduction
Y (s) 1 Modelling of
H(s) = = dynamical
U(s) s(s + 1) systems
Properties
Adding the Zero order holder leads to: Discrete-time
systems
Conclusion
hence
1 1 1
Z [GBOZ (s)H(s)] = (1 − z −1 )Z 2
− +
s s s+1
State space
Exercise (cont’d) representations
(SEM)
O.Sename
1 1 1 Introduction
Z [GBOZ (s)H(s)] = (1 − z −1 )Z 2
− +
s s s+1 Modelling of
dynamical
zT e z z systems
= (1 − z −1 ) 2
− + −T Properties
(z − 1) z −1 z −e e
Discrete-time
(ze−Te − z + zTe ) + (1 − e−Te − Te e−Te ) systems
=
(z − 1)(z − e−Te ) State feedback
control
Observer
if Te = 1, we have Integral Control
A polynomial
(ze−Te − z + zTe ) + (1 − e−Te − T e e−Te ) approach
Z [GBOZ (s)H(s)] =
(z − 1)(z − e−Te ) Further in
discrete-time
ze−1 + 1 − 2e−1 control
= Conclusion
(z − 1)(z − e−1 )
b1 z + b0
= 2
z + a1 z + a0
State space
Exercise (cont’d) representations
(SEM)
Y (z) b1 z+b0
U(z) = z 2 +a1 z+a0
Introduction
b1 z+b0 Modelling of
⇔ Y (z) = z 2 +a1 z+a0
U(z) dynamical
systems
⇔ 2
Y (z)(z + a1 z + a0 ) = (b1 z + b0 )U(z) Properties
⇔ y(n + 2) + a1y(n + 1) + a0y(n) = b1 u(n + 1) + b0u(n) Discrete-time
systems
With an unit feedback, the closed loop function is given by: State feedback
control
G(z)
Fcl (z) = Observer
1 + G(z) Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
State feedback
Stability control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Equivalence {s} ↔ {z} representations
(SEM)
z = esTe Properties
Discrete-time
State feedback
the same pole in Z. control
Observer
Stability domain Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Approximations representations
(SEM)
O.Sename
Introduction
Forward difference (Rectangle inferior) Modelling of
dynamical
systems
Properties
z −1 Discrete-time
s= systems
Te
State feedback
control
Observer
Backward difference (Rectangle superior) Integral Control
A polynomial
approach
Further in
z −1 discrete-time
s= control
zTe Conclusion
State space
Approximations (cont’d) representations
(SEM)
Introduction
Modelling of
dynamical
2 z −1 systems
s=
Te z + 1 Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Systems definition representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
A discrete-time state space system is as follows: systems
Properties
(
x((k + 1)h) = Ad x(kh) + Bd u(kh), x(0) = x0 Discrete-time
systems
(22)
y(kh) = Cd x(kh) + Dd u(kh) State feedback
control
Observer
where h is the sampling period.
Integral Control
Matlab : ss(Ad ,Bd ,Cd ,Dd ,h) creates a SS A polynomial
object SYS representing a discrete-time approach
Conclusion
State space
Relation with transfer function representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
For discrete-time systems, systems
Properties
x((k + 1)h) = Ad x(kh) + Bd u(kh), x(0) = x0 Discrete-time
(23) systems
y(kh) = Cd x(kh) + Dd u(kh)
State feedback
control
the discrete transfer function is given by Observer
Integral Control
G(z) = Cd (zIn − Ad )−1 Bd + Dd (24) A polynomial
approach
Conclusion
State space
Recall Laplace & Z-transform representations
(SEM)
Introduction
Modelling of
dynamical
systems
H(s) to state space H(z) to state space
Properties
Discrete-time
systems
X X
U = den(s) U = den(z) State feedback
Y Y control
X = num(s) X = num(z)
Observer
Integral Control
Introduction
The state xk , solution of system xk +1 = Ad xk with initial Modelling of
condition x0 , is given by dynamical
systems
Properties
x1 = Ad x0 (25)
Discrete-time
x2 = A2d x0 (26) systems
State feedback
xn = And x0 (27) control
Observer
A polynomial
approach
x1 = Ad x0 + Bd u0 (28)
Further in
x2 = A2d x0 + Ad Bd u0 + Bd u1 (29) discrete-time
control
n−1 Conclusion
xn = And x0 + ∑ An−1−i
d Bd ui (30)
i=0
State space
State space analysis (discrete-time systems) representations
(SEM)
Stability O.Sename
Discrete-time
Definition systems
Given two states x0 and x1 , the system (22) is controllable State feedback
control
if there exist K1 > 0 and a sequence of control samples
Observer
u0 , u1 , . . . , uK1 , such that xk takes the values x0 for k = 0
Integral Control
and x1 for k = K1 . A polynomial
approach
O.Sename
Controllability
Introduction
The system is controllable iff
Modelling of
dynamical
C⌈ (A = rg[Bd Ad Bd . . . An−1
d Bd ] =n systems
d ,Bd )
Properties
Discrete-time
systems
Observability State feedback
control
The system is observable iff
Observer
Integral Control
O(Ad ,Cd ) = rg[Cd Cd Ad . . . Cd An−1 T
d ] =n A polynomial
approach
Further in
discrete-time
Duality control
(proof. . . )
Controllability of (Ad , Bd ) ⇔ Observability of (BdT , ATd ).
(proof. . . )
State space
State feedback representations
(SEM)
O.Sename
Introduction
A state feedback controller for a continuous-time system is:
Modelling of
dynamical
u(t) = −Fx(t) (31) systems
Properties
Observer
When the system is MIMO we have
Integral Control
A polynomial
u1 x1 approach
u2 f11 . . . f1n
.. ..
x2
Further in
.. = . . .. discrete-time
control
. .
fm1 . . . fmn Conclusion
um xn
State space
State feedback (2) representations
(SEM)
O.Sename
Introduction
Using state feedback controllers (31), we get in Modelling of
dynamical
closed-loop (for simplicity D = 0) systems
Properties
ẋ(t) = (A − BF )x(t), Discrete-time
(32)
y(t) = Cx(t) systems
State feedback
control
and the stability (and dynamics) of the closed-loop system Observer
is then given by the eigenvalues of A − BF . Integral Control
For discrete-time system we get: A polynomial
approach
x(k + 1) = (A − BF )x(k), Further in
(33) discrete-time
y(k) = Cx(k) control
Conclusion
State space
State feedback (3) representations
(SEM)
Modelling of
u(t) = −Fx(t) + Gr (t) (34) dynamical
systems
or u(k) = −Fx(k) + Gr (k) (35) Properties
Discrete-time
G is a m × p real matrix. Then the closed-loop transfer systems
Observer
GCL (s) = C(sIn − A + BF )−1BG (36) Integral Control
A polynomial
G is chosen to ensure a unitary steady-state gain as: approach
Further in
discrete-time
G = [C(−A + BF )−1B]−1 (37) control
Conclusion
⋆ For discrete-time system:
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Proposition Discrete-time
Let a linear system given by A, B, and let γi , i = 1, ..., n , a systems
State feedback
set of complex elements (i.e. the desired poles of the control
closed-loop system). There exists a state feedback control Observer
u = −Fx such that the poles of the closed-loop system are Integral Control
Further in
discrete-time
control
Conclusion
State space
Why state feedback and not output feedback? representations
(SEM)
O.Sename
Introduction
1 Modelling of
Example: G(s) = s2 −s dynamical
systems
Consider the canonical form.
Properties
Case of output feedback : u = −Ly
Discrete-time
Then ẋ(t) = (A − BLC)x(t) systems
Further in
So we can choose any F . For instance f1 = 1, f2 = 3 gives discrete-time
Conclusion
State space
Pole placement control (1) representations
(SEM)
0 1 0 ... 0 0 Introduction
0 0 1 0 ... ..
..
. Modelling of
.. .. . . .. dynamical
A= . . . . . , B= .. and systems
.
.. Properties
0 . 0 1 0
Discrete-time
−a0 −a1 . . . . . . −an−1 1 systems
State feedback
C= c0 c1 ... cn−1 . control
Let F = [ f1 f2 . . . fn ] Observer
A polynomial
0 1 0 ... 0 approach
0 0 1 0 ...
Further in
discrete-time
.
.. .
.. .. .. .. control
A − BF = . . .
Conclusion
..
0 . 0 1
−a0 − f1 −a1 − f2 . . . . . . −an−1 − fn
(39)
State space
Pole placement control (2) representations
(SEM)
O.Sename
Introduction
Discrete-time
systems
fi = −ai−1 + αi−1 , i = 1, .., n State feedback
control
ensures that the poles of A − BF are {γi }, i = 1, n Observer
O.Sename
Introduction
Modelling of
Problem: To implement a state feedback control, the dynamical
systems
measurement of all the state variables is necessary. If this Properties
is not available, we will use a state estimation through a Discrete-time
so-called Observer. systems
Observer
˙
x̂(t) = Ax̂(t) + Bu(t) − L(C x̂(t) − y(t))
(40) Integral Control
x̂0 A polynomial
approach
where x̂(t) ∈Rn is the estimated state of x(t) and L is the Further in
discrete-time
n × p constant observer gain matrix to be designed. control
Conclusion
State space
Observer representations
(SEM)
O.Sename
Th estimated error, e(t) := x(t) − x̂(t), satisfies:
Introduction
ė(t) = (A − LC)e(t) (41) Modelling of
dynamical
systems
If L is designed such that A − LC is stable, then x̂ (t) Properties
converges asymptotically towards x(t). Discrete-time
systems
Proposition State feedback
control
(40) is an observer for system (2) if and only if the pair
Observer
(C,A) is observable, i.e. Integral Control
A polynomial
rank(O) = n approach
Further in
discrete-time
C control
CA
Conclusion
where O = .. .
.
CAn−1
State space
Observer design representations
(SEM)
In order to use the acker Matlab function, we will use the Modelling of
dynamical
duality property between observability and controllability, systems
i.e. : Properties
Integral Control
stable.
A polynomial
Matlab : use L=acker(A’,C’,Po)’ where Po approach
O.Sename
Modelling of
dynamical
u(t) = −F x̂(t) + Gr (t) (42) systems
Properties
We then need to study the stability of the complete Discrete-time
systems
closed-loop system, using the extended state:
State feedback
T control
xe (t) = x(t) e(t) Observer
Integral Control
The closed-loop system with observer (40) and control A polynomial
approach
(42) is:
Further in
discrete-time
A − BF BF BG control
ẋe (t) = xe (t) + r (t) (43)
0 A − LC 0 Conclusion
State space
Separation principle representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
The characteristic polynomial of the extended system is: systems
Properties
det(sIn − A + BF ) × det(sIn − A + LC) Discrete-time
systems
State feedback
If the observer and the control are designed separately control
then the closed-loop system with the dynamic Observer
measurement feedback is stable, given that the control and Integral Control
Conclusion
State space
Stabilisation/ Detectability representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
When the linear system (2) is not completely controllable systems
or observable, it is then important to study the stability of State feedback
control
the non controllable and non observable modes.
Observer
Use ctrbf and obsvf Matlab commands Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Integral Control representations
(SEM)
O.Sename
Introduction
A state feedback controller may not allow to reject the Modelling of
effects of disturbances (particularly of input disturbances). dynamical
systems
A very useful method consists in adding an integral term to Properties
ensure a unitary static closed-loop gain . Discrete-time
systems
Considered system:
State feedback
( control
ẋ(t) = Ax(t) + Bu(t) + Ed (t), x(0) = x0 Observer
(44)
y(t) = Cx(t) Integral Control
A polynomial
approach
where d is the disturbance. Further in
The objective is to keep y close to a reference signal r , discrete-time
control
even in the presence of d , i.e to keep r − y asymptotically Conclusion
stable.
State space
Integral Control representations
(SEM)
O.Sename
Introduction
The method consists in extending the system by adding a Modelling of
new state variable: dynamical
systems
Properties
ż(t) = r (t) − y(t) Discrete-time
systems
and to use a new state feedback: State feedback
control
Observer
u(t) = −Fx(t) − Hz(t)
Integral Control
We get A polynomial
approach
Further in
ẋ(t) A − BF BH x 0 E discrete-time
= + r (t) + d (t) control
ż(t) −C 0 z 1 0
Conclusion
State space
Integral control scheme representations
(SEM)
O.Sename
Introduction
The complete structure has the following form: Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Introduction
Modelling of
dynamical
The use of an observer-based controller is equivalent to systems
Discrete-time
systems
−1
u(s) = −F (sIn − A + BF + LC) Ly(s) State feedback
control
−1
+[In − F (sIn − A + BF + LC) B]Gr (s) Observer
Integral Control
which corresponds to a two-degrees of freedom controller
A polynomial
approach
R(s) T (s)
u(s) = − y(s) + r (s) Further in
discrete-time
S(s) S(s) control
O.Sename
Influence of the sampling period on the time response
Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
Impose a maximal time response to a discrete system is control
equivalent to place the poles inside a circle defined by the Conclusion
used.
Introduction
Example with sampling Time Modelling of
Te = 1s ⇔ fe = 1Hz ⇔ we = 2π ): dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
Note that, in our case, the Bode is cut at the pulse w = π . control
used.
Introduction
Example with sampling Time Modelling of
Te = 1s ⇔ fe = 1Hz ⇔ we = 2π ): dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
Note that, in our case, the Bode is cut at the pulse w = π . control
Sampling ↔ Limitations
Recall the Shannon theorem that impose the sampling
frequency at least 2 times higher that the system
State space
About sampling period and robustness representations
(SEM)
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
But reducing the sampling time modify poles discrete-time
control
location. . . Poles and zeros become closer to the limit of
Conclusion
the unit circle ⇒ can introduce instability (decrease
robustness).
⇒ Sampling influences stability and robustness
⇒ Over sampling increase noise sensitivity
State space
Zeros representations
(SEM)
O.Sename
Properties
zero at the origin induces a sample advance.
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Stability representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Recall Properties
A linear continuous feedback control system is stable if all Discrete-time
poles of the closed-loop transfer function T (s) lie in the left systems
A polynomial
|z| = eσ Te and ∠z = ω Te approach
Further in
discrete-time
control
Conclusion
State space
Stability (cont’d) representations
(SEM)
O.Sename
Introduction
Jury criteria
Modelling of
The denominator polynomial dynamical
systems
(den(z) = a0 z n + a1 z n−1 + · · · + an = 0) has all its roots Properties
inside the unit circle if all the first coefficients of the odd Discrete-time
row are positive. systems
State feedback
an control
b = a0 − an Observer
1 a0 a1 a2 ... an−k . . . an 0 a0
Integral Control
2 an an−1 an−2 ... ak ... a0 an
b1 = a1 − an−1 A polynomial
3 b0 b1 b2 ... bn−1 aapproach
0
2 bn−1 bn−2 bn−3 ... b0 aFurther
n in
.. .. bk = ak − an−k discrete-time
. . acontrol
0
2n + 1 b
Conclusion
s0 ck = bk − bn−1−k n−1
b0
State space
Example representations
(SEM)
O.Sename
Stability
Find the stability region of D(z) = z 2 + a1 z + a2 Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Example representations
(SEM)
O.Sename
Stability
Find the stability region of D(z) = z 2 + a1 z + a2 Introduction
Modelling of
dynamical
Solution systems
Properties
Discrete-time
systems
1 1>0 a1 a2
State feedback
2 a2 a1 1 control
Further in
discrete-time
hence, control
Conclusion
1 − a22 > 0
(1 + a2)2 > a21
State space
How to get a discrete controller representations
(SEM)
O.Sename
First way
Introduction
State feedback
Second way control
Observer
◮ Design a continuous-time controller Integral Control
Conclusion
Now the question is how to implement the computed
controller on a real-time (embedded) system, and what are
the precautions to take before?
State space
representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
Implementation Discrete-time
systems
State feedback
characteristics control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Anti-aliasing & Sampling representations
(SEM)
Anti-aliasing O.Sename
Properties
to the frequency used by the controller. Remember that
Discrete-time
the pre-filter introduce phase shift. systems
State feedback
Sampling frequency choice control
Integral Control
the desired speed of the closed loop system. A rule of
A polynomial
thumb is that one should sample 4 − 10 times per rise time approach
Tr of the closed loop system. Further in
discrete-time
control
Tr
Nsample = ≈ 4 − 10 Conclusion
Te
where Te is the sampling period, and Nsample the number
of samples.
State space
Delay representations
(SEM)
O.Sename
Problematic
Introduction
Sampled theory assume presence of clock that
Modelling of
synchronizes all measurements and control signal. Hence dynamical
systems
in a computer based control there always is delays (control
Properties
delay, computational delay, I/O latency). Discrete-time
systems
Origins State feedback
control
There are several reasons for delay apparition
Observer
◮ Execution time (code) Integral Control
Conclusion
◮ Data dependencies
Hence the control delay is not constant. The delay
introduce a phase shift ⇒ Instability!
State space
Delay (cont’d) representations
(SEM)
PM π State feedback
DM = [s] control
180w0 Observer
Integral Control
A polynomial
Exercise: compute delay margin for these 3 cases approach
Further in
discrete-time
control
Conclusion
State space
Delay (cont’d) representations
(SEM)
O.Sename
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Delay (cont’d) representations
(SEM)
Modelling of
◮ Minimize the delay (case B - Minimal control delay) dynamical
systems
◮ Compensate it off-line Properties
◮ Make the controller robust (case A - static scheduling) Discrete-time
systems
◮ Compensate on-line State feedback
control
Integral Control
LOOP A polynomial
%%% At each clock interrup
approach
ADin Further in
discrete-time
CalculateOutput control
DAout Conclusion
UpdateStates
IncTime %%% Evaluate remaining tim
WaitUntilTe
END
State space
Delay (cont’d) representations
(SEM)
Exercise O.Sename
Modelling of
x(k + 1) = Fx(k) + Gy(k) dynamical
systems
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Delay (cont’d) representations
(SEM)
Exercise O.Sename
Modelling of
x(k + 1) = Fx(k) + Gy(k) dynamical
systems
Discrete-time
systems
State feedback
control
Observer
LOOP
Integral Control
ADin(y); A polynomial
%%% CalculateOutput approach
u := u1 + D*y; Further in
discrete-time
DAout(u) Note that such a structure is not
control
Effects O.Sename
Modelling of
◮ Limit cycles dynamical
systems
Properties
Example (stable for K<2) Discrete-time
systems
State feedback
control
0.25
H(z) = Observer
(z − 1)(z − 0.5) Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Quantification (cont’d) representations
(SEM)
O.Sename
Results Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Quantification (cont’d) representations
(SEM)
O.Sename
Results Introduction
Modelling of
dynamical
systems
Properties
Discrete-time
systems
State feedback
control
Observer
Integral Control
A polynomial
approach
Further in
discrete-time
control
Conclusion
State space
Discretisation representations
(SEM)
O.Sename
Discrete-time
To obtain this, few methods exists that approach the systems
Recall Observer
Integral Control
A polynomial
z −1 approach
s = Further in
Te discrete-time
control
z −1
s = Conclusion
zTe
2 z −1
s =
Te z + 1
State space
Conclusion representations
(SEM)
O.Sename
Introduction
Modelling of
dynamical
systems
Properties
◮ A state space approach to pole placement control Discrete-time
systems
◮ A similar design can be done using a polynomial
State feedback
approach control
Observer
◮ Continuous but directly extended to discrete-time
Integral Control
systems.
A polynomial
approach
Further in
discrete-time
control
Conclusion