100% found this document useful (1 vote)
102 views

Module 11: Introduction To Optimal Control: Lecture Note 3

The document discusses the linear quadratic regulator problem for finding an optimal controller. It describes the system model and performance index to be minimized. It then derives the discrete algebraic Riccati equation and uses it to find the optimal controller that minimizes a given quadratic performance index for a example system.

Uploaded by

kosu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (1 vote)
102 views

Module 11: Introduction To Optimal Control: Lecture Note 3

The document discusses the linear quadratic regulator problem for finding an optimal controller. It describes the system model and performance index to be minimized. It then derives the discrete algebraic Riccati equation and uses it to find the optimal controller that minimizes a given quadratic performance index for a example system.

Uploaded by

kosu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Digital Control

Module 11

Lecture 3

Module 11: Introduction to Optimal Control


Lecture Note 3

Linear Quadratic Regulator

Consider a linear system modeled by


x(k + 1) = Ax(k) + Bu(k),

x(k0 ) = x0

where x(k) Rn and u(k) Rm . The pair (A, B) is controllable.


The objective is to design a stabilizing linear state feedback controller u(k) = Kx(k) which
will minimize the quadratic performance index, given by,
J=

(xT (k)Qx(k) + uT (k)Ru(k))

k=0

where, Q = QT 0 and R = RT > 0. Such a controller is denoted by u .


We first assume that a linear state feedback optimal controller exists such that the closed
loop system
x(k + 1) = (A BK)x(k)
is asymptotically stable.
This assumption implies that there exists a Lyapunov function V (x(k)) = x(k)T P x(k) for
the closed loop system, for which the forward difference
V (x(k)) = V (x(k + 1)) V (x(k))
is negative definite.
We will now use the theorem as discussed in the previous lecture which says if the controller u
is optimal, then
min(V (x(k)) + xT (k)Qx(k) + uT (k)Ru(k)) = 0
u

I. Kar

Digital Control

Module 11

Lecture 3

Now, finding an optimal controller implies that we have to find an appropriate Lyapunov function which is then used to construct the optimal controller.
Let us first find the u that minimizes the function
f = f (u(k)) = V (x(k)) + xT (k)Qx(k) + uT (k)Ru(k)
If we substitute V in the above expression, we get
f (u(k)) = xT (k + 1)P x(k + 1) xT (k)P x(k) + xT (k)Qx(k) + uT (k)Ru(k)
= (Ax(k) + Bu(k))T P (Ax(k) + Bu(k)) xT (k)P x(k) + xT (k)Qx(k) + uT (k)Ru(k)
Taking derivative of the above function with respect to u(k),
f (u(k))
= 2(Ax(k) + Bu(k))T P B + 2uT R
u(k)
= 2xT (k)AT P B + 2uT (k)(B T P B + R)
= 0T
The matrix B T P B + R is positive definite since R is positive definite, thus it is invertible.
Hence,
u (k) = (B T P B + R)1 B T P Ax(k) = Kx(k)
where K = (B T P B + R)1 B T P A. Let us denote B T P B + R by S. Thus
u (k) = S 1 B T P Ax(k)
We will now check whether or not u satisfies the second order sufficient condition for minimization. Since
2 f (u(k))

=
(2xT (k)AT P B + 2uT (k)(B T P B + R))
u2 (k)
u(k)
= 2(B T P B + R) > 0
u satisfies the second order sufficient condition to minimize f .

I. Kar

Digital Control

Module 11

Lecture 3

The optimal controller can thus be constructed if an appropriate Lyapunov matrix P is found.
For that let us first find the closed loop system after introduction of the optimal controller.
x(k + 1) = (A BS 1 B T P A)x(k)
Since the controller satisfies the hypothesis of the theorem, discussed in the previous lecture,
xT (k + 1)P x(k + 1) xT (k)P x(k) + xT (k)Qx(k) + u T (k)Ru (k) = 0
Putting the expression of u in the above equation,
xT (k)(A BS 1 B T P A)T P (A BS 1 B T P A)x(k) xT (k)P x(k) + xT (k)Qx(k) +
xT (k)AT P BS 1 RS 1 B T P Ax(k)
= xT (k)AT P Ax(k) xT (k)AT P BS 1 B T P Ax(k) xT (k)AT P BS 1 B T P Ax(k) +
xT (k)AT P BS 1 B T P BS 1 B T P Ax(k) xT (k)P x(k) + xT (k)Qx(k) +
xT (k)AT P BS 1 RS 1 B T P Ax(k)
= xT (k)AT P Ax(k) xT (k)P x(k) + xT (k)Qx(k) 2xT (k)AT P BS 1 B T P Ax(k) +
xT (k)AT P BS 1 (B T P B + R)S 1 B T P Ax(k)
= xT (k)AT P Ax(k) xT (k)P x(k) + xT (k)Qx(k)
2xT (k)AT P BS 1 SS 1 B T P Ax(k) + xT (k)AT P BS 1 B T P Ax(k)
= xT (k)(AT P A P + Q AT P BS 1 B T P A)x(k) = 0
The above equation should hold for any value of x(k). Thus
AT P A P + Q AT P BS 1 B T P A = 0
which is the well known discrete Algebraic Riccati Equation (ARE). By solving this equation
we can get P to form the optimal regulator to minimize a given quadratic performance index.
Example 1: Consider the following linear system

x(k + 1) =
y(k) =





 
0
0
1
u(k),
x(k) +
1
0.5 0.8

 
1
x0 =
1


1 0 x(k)

Design an optimal controller to minimize the following performance index.


J=

(x21 + x1 x2 + x22 + 0.1u2 )

k=0

I. Kar

Digital Control

Module 11

Lecture 3

Also, find the optimal cost.


Solution: The performance index J can be rewritten as
J=

X
k=0

Thus, Q =


1 0.5
x(k) + 0.1u2 )
(x (k)
0.5 1
T

1 0.5
and R = 0.1.
0.5 1

Let us take P as


p p
P = 1 2
p2 p3

Then,


0.25p3 p1
0.5p2 + 0.4p3 p2
A PA P =
0.5p2 + 0.4p3 p2 p1 + 1.6p2 + 0.64p3 p3
T

0.25p3 p1 + 1
0.4p3 0.5p2 + 0.5
A PA P + Q =
0.4p3 0.5p2 + 0.5 p1 + 1.6p2 0.36p3 + 1
T


0.5p3
,
A PB =
p2 + 0.8p3


A P BS



B T P A = 0.5p3 p2 + 0.8p3 ,

S = 0.1 + p3





1
0.5p3
0.5p3 p2 + 0.8p3
B PA = =
0.1 + p3 p2 + 0.8p3


1
0.25p23
0.5p2 p3 + 0.4p23
=
0.1 + p3 0.5p2 p3 + 0.4p23 p22 + 1.6p2 p3 + 0.64p23
T

The discrete ARE is


AT P A P + Q AT P BS 1 B T P A = 0
Or,

0.25p3 p1 + 1
0.4p3 0.5p2 + 0.5

I. Kar

0.25p23
0.1+p3

0.5p2 p3 +0.4p23
0.1+p3



0.5p2 p3 + 0.4p23
0.4p3 0.5p2 + 0.5
0 0

0.1 + p3
=
0 0
p22 +1.6p2 p3 +0.64p23
p1 + 1.6p2 0.36p3 + 1
0.1+p3

Digital Control

Module 11

Lecture 3

We can get three equations from the discrete ARE. These are
0.25p3 p1 + 1

0.4p3 0.5p2 + 0.5

p1 + 1.6p2 0.36p3 + 1

0.25p23
=0
0.1 + p3

0.5p2 p3 + 0.4p23
=0
0.1 + p3

p22 + 1.6p2 p3 + 0.64p23


=0
0.1 + p3

Since the above three equations comprises three unknown parameters, these parameters can be
solved uniquely, as
p1 = 1.0238,

p2 = 0.5513,

p3 = 1.9811

The optimal control law can be found out as


u (k) = (R + B T P B)1 B T P Ax(k)
= [0.4760 1.0265]x(k)
= 0.4760x1 (k) 1.0265x2 (k)
The optimal cost can be found as
J = xT0 P x0
 


 1.0238 0.5513 1
= 1 1
0.5513 1.9811 1
= 4.1075

I. Kar

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy