New Approaches to the Design

of Fixed Order Controllers

S. P. Bhattacharyya
Department of Electrical Engineering
Texas A & M University
College Station, TX 77843-3128
Main highlights

New algorithms that can be used to generate the entire set of stabilizing PID and first order
controllers for single-input single-output 1) continuous-time rational plants of arbitrary
order, 2) discrete-time rational plants of arbitrary order, and 3) continuous-time plants with
time delay. A new linear programming algorithm for higher order controllers.

These algorithms follow from substantial fundamental theoretical advances on PID, first
order and arbitrary fixed order stabilization that have been reported by us in recent years.
They display the rich mathematical structure underlying the topology of stabilizing sets.

Examples are presented to clarify the steps involved in implementing the different
algorithms.

The results significantly complement the current techniques for industrial PID design, many
of which are adhoc in nature, and open the door to design with fixed order controllers.

Our solution results in graphical displays of feasible design regions using 2-D and 3-D
graphics-should appeal to control designers and are very suitable for computer aided design
where several performance objectives have to be overlaid and intersected.

Motivation

Fixed order and in particular PID Controllers are widely used in aerospace, motion control,
process industries, manufacturing, robotics, disc drives, pneumatic, hydraulic, electrical and
mechanical systems

Current designs are carried out using adhoc tuning rules. These rules have been developed
over the years based primarily on empirical observations and industrial experience.

Modern and Postmodern control theory, that is, state feedback observer based theory of
modern and post-modern control theory including optimal control cannot be applied to fixed
order controllers. Indeed, until the results to be described here appeared, it was not known
how to even determine whether stabilization of a nominal system was possible using PID
controllers.

Our results are first steps toward computer aided designs of fixed order and PID controllers
with guaranteed stability, performance and robustness.
2
H H

, ,
1
l
Characteristics of PID Controllers





Provides set point regulation (error zeroing) against arbitrary
disturbances (as long as they are low frequency)

Is robust against modelling errors

Is nonfragile in general

Three term controllers are easier to adjust at the design stage as
well as online
Theoretical developments




Extension of the classical Hermite Bieler Theorem to root
counting

A new Stabilization Algorithm based on this, that generates
the entire set of controllers attaining stability

Calculations based on Linear Programming

A new Performance Algorithm for Robustness, Gain-Phase
Margin and other performance measures generating the entire
set attaining specs.
H

PID Controllers for Linear Time-invariant

Continuous-time Systems
Figure 1. Feedback control system.

( )
i
p d
k
C s k k s
s
= + +
where and are the proportional, integral and
derivative gains respectively. Plant transfer function

,
p
k
i
k
d
k
( )
( )
( )
N s
G s
D s
=
where , are polynomials in the Laplace variable .
The closed-loop characteristic polynomial is

( ) N s ( ) D s s
2
( ) ( ) ( ) ( ) ( )
p i d i d p
s k k k sD s k k s N s k sN s o , , , = + + + .
Stabilization Problem: Determine the values of and
for which the closed-loop characteristic polynomial
is Hurwitz, that is, has all its roots in the open left half plane.
,
p
k
i
k
d
k
( )
p i d
s k k k o , , ,
Some notation:
The standard signum function is defined by
{ 1 0 1} sgn R : , ,
1 if 0
[ ] 0 if 0
1 if 0
x
sgn x x
x
<

= =

> .

0 1
( )
n
n
a s a a s a s = + + +
Then and denote the numbers of roots of
in the open LHP and RHP respectively
( ( )) l a s ( ( )) r a s ( ) a s
The even-odd decomposition of is defined as ( ) a s
2 2
( ) ( ) ( )
e o
a s a s sa s = +
where and are the components of made up
of even and odd powers of respectively.
2
( )
e
a s
2
( )
o
sa s ( ) a s
s
2 2
( ) ( ) ( )
e o
a j a j a e e e e = + .
Basic Idea: Can determine the root distribution of from
knowledge of the the zeros of the odd part and the signs of the
even part evaluated at these zeros.

( ) a s
Using the even-odd decomposition of , define ( ) N s
( ) ( ) ( ) ( )
2 2
e o
N s N s N s sN s
-
= =
Let , be the degrees of and respectively.
To achieve parameter separation multiply by
to obtain

n m
( )
, , ,
p i d
s k k k o
( ) N s
( )
, , ,
p i d
s k k k o ( ) N s
-
( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )
2 2 2 2 2
: , , ,
p i d e o e o
v s s k k k N s s N s D s D s N s o
-

= =

( ) ( ) ( ) ( ) ( ) ( )
2 2 2 2 2 2
i d e e o o
k k s N s N s s N s N s
(
+ +

( ) ( ) ( ) ( )
2 2 2 2 2
e e o o
s D s N s s D s N s

( ) ( ) ( ) ( ) ( )
2 2 2 2 2
p e e o o
k N s N s s N s N s
(
+

is Hurwitz if and only if has exactly the
same number of closed RHP zeros as
( )
p i d
s k k k o , , , ( ) s v
( ) N s
-
has all three parameters appearing in both the
even and odd parts, but the test polynomial exhibits
parameter separation, that is, appears in the odd part only
while and appear in the even part only. This facilitates
the application of root counting formulas to .
( )
p i d
s k k k o , , ,
( ) s v
p
k
i
k
d
k
( ) s v
( ) ( ) ( ) ( )
p i d i d p
j k k k N j p k k jq k o e e e e
-
, , , = , , + ,
where
2
1 2
1 2
2 2 2 2 2
1
2 2 2 2 2
2
2 2 2 2 2
1
2 2 2
2
( ) ( ) ( ) ( )
( ) ( ) ( )
( ) ( ( ) ( ) ( ) ( ))
( ) ( ) ( ) ( ) ( ))
( ) ( ( ) ( ) ( ) ( ))
( ) ( ( ) ( ) (
i d i d
p p
e o e o
e e o o
e e o o
e e o
p k k p k k p
q k q k q
p N D D N
p N N N N
q D N D N
q N N N
e e e e
e e e
e e e e e e
e e e e e e
e e e e e e e
e e e e e e
, , = +
, = +
=
= +
= +
= +
2 2
) ( ))
o
N e .
For stability must have exactly the same number of
RHP roots as . For this a necessary condition is that
has at least
( ) s v
( ) N s
( )
p
q k e,
( ( ( )) ( ( )))
for even
2
( ( ( )) ( ( ))) 1
for odd
2
n l N s r N s
m n
n l N s r N s
m n

+

real, nonnegative, distinct roots of odd multiplicity. The

ranges of satisfying this condition are called allowable.
Let
p
k
0 1 1
0
l
e e e

= < < ... <
denote the real, nonnegative distinct roots of of odd
multiplicity, and with write

( )
p
q k e,
l
e :=
[ ( )] 0 1
j j
sgn p i j l e = , = , , .... .
It can be shown using the root counting results mentioned
earlier, that the stability condition reduces to:
1
0 1 2 1
1
1
0 1 2 1
1
{ 2 2 ( 1) 2
( 1) }( 1) sgn[ ( )]
for even
( ( ( )) ( ( )))
{ 2 2 ( 1) 2 }
( 1) sgn[ ( )]
for odd
l
l
l l
l p
l
l
l
p
i i i i
i q k
m n
n l N s r N s
i i i i
q k
m n

+ + +

+ ,

+ + +

and therefore the string of integers will be called

admissible if it satisfies the above condition
0 1
{ }
l
i i i , , ...,
PID Stabilization Algorithm For
LTI Plants:
Step 1: For the given and , compute the
corresponding , , , and

( ) N s ( ) D s
1
( ) p e
2
( ) p e
1
( ) q e
2
( ) q e
Step 2: Determine the allowable ranges of .
The resulting ranges of are the only ranges of for which
stabilizing values may exist;
1 2
i
P i d , = , , ,
p
k
p
k
p
k
( )
i d
k k ,
Step 3: If there is no satisfying Step 2 then output NO
SOLUTION and EXIT;
p
k
Step 4: Initialize and ; 1 j =
j
P P =
Step 5: Pick a range in and initialize ;
[ ]
low upp
k k ,
P
p low
k k =
Step 6: Pick the number of grid points and set ; N
1
1
[ ]
upp low N
step k k
+
=
Step 7: Increase as follows: . If
then GOTO Step 14;

p
k
p p
k k step = +
p upp
k k >
Step 8: For fixed in Step 7, solve for the real, non-
negative, distinct finite zeros of with odd
multiplicities and denote them by .
Also define ;

p
k
( )
p
q k e,
0 1 2 1
0
l
e e e e

= < < < <
l
e =
Step 9: Construct sequences of numbers as follows:
0 1 2 l
i i i i , , , ,
(i) If for some , then define ( ) 0
t
N je
-
= 1 2 1 t l = , , ,
0
t
i = ;
(ii) For all other , 0 1 2 t l = , , , ,
{ 1 1}
t
i e , .
With defined in this way, define the set as
0 1
i i , ,
( )
p
k
A
0 1
( )
0 1 1
{{ }} if is even
{{ }} if is odd
p
l
k
l
i i i m n
A
i i i m n

, , , +
:=

, , , + ;

Step 10: Determine the admissible strings in

from (??). If there is no admissible string then GOTO Step
7;

0 1
{ } I i i = , ,
( )
p
k
A
Step 11: For an admissible string , determine the
set of values that simultaneously satisfy the following
string of linear inequalities:
{ }
0 1
I i i = , ,
( )
i d
k k ,
2
1 2
[ ( ) ( ) ( )] 0 0 1 2
t i d t t t
p k k p i t e e e + > , = , , ,
Step 12: Repeat Step 11 for all admissible strings
to obtain the corresponding admissible sets .
The set of all stabilizing values corresponding to the
fixed is then given by
1 2 v
I I I , , ,
( )
i d
k k ,
1 2 v
S S S , , , ( )
i d
k k ,
p
k
( ) 1 2
p
k x v x
S S
= , , ,
= ;
Step 13: GOTO Step 7
Step 14: Set and . If GOTO STEP 5;
else, terminate the algorithm.
1 j j = +
j
P P = j d s
Example
Consider the problem of determining stabilizing PID gains for the
plant where
The closed-loop characteristic polynomial is
Thus and . Also
and
Therefore we obtain
so that
where
( )
( )
( )
N s
D s
G s =
3 2
6 5 4 3 2
( ) 2 1
( ) 2 32 26 65 8 1
N s s s s
D s s s s s s s
=
= + + + + + .
2
( ) ( ) ( ) ( ) ( )
p i d i d p
s k k k sD s k k s N s k sN s o , , , = + + + .
7 n = 3 m=
2 2 2 2 2 6 4 2 2 4 2
( ) 2 1 ( ) 1 ( ) 32 65 1 ( ) 2 26 8
e o e o
N s s N s s D s s s s D s s s = , = , = + + + , = + ,
2 2
( ) ( 2 1) ( 1) N s s s s
-
= .



( ) ( )
p i d
s k k k N s o
-
, , , =
2 8 6 4 2 2
[ ( 35 87 54 9) ( )
i d
s s s s s k k s + + + +
6 4 2 8 6 4
( 6 3 1)] [( 4 89 128 s s s s s s s + + + +
2 6 4 2
75 1) ( 6 3 1)]
p
s k s s s + + + +
2
1 2 1 2
( ) ( ) [ ( ) ( ) ( )] [ ( ) ( )]
p i d i d p
j k k k N j p k k p j q k q o e e e e e e e
-
, , , = + + +
10 8 6 4 2
1
6 4 2
2
9 7 5 3
1
7 5 3
2
( ) 35 87 54 9
( ) 6 3 1
( ) 4 89 128 75
( ) 6 3
p
p
q
q
e e e e e e
e e e e
e e e e e e
e e e e e
= + +
= + +
= + +
= + + .
In Step 2, the range of such that has at least real,
non-negative, distinct, finite zeros with odd multiplicities was
determined to be which is the allowable range.
Now for a fixed , for instance , we
have

p
k ( )
f p
q k e,
3
( 24 7513 1) . ,
( 24 7513 1)
p
k e . , 18
p
k =
1 2
9 7 5 3
( 18) ( ) 18 ( )
4 71 236 129 19
q q q e e e
e e e e e
, =
= + + .
Then the real, non-negative, distinct finite zeros of
with odd multiplicities are
( 18)
f
q e,
0 1 2 3 4
0 0 5195 0 6055 1 8804 3 6848 e e e e e = , = . , = . , = . , = . .
Also define . Since which is even, and
and ,
5
e = 10 m n + = ( ( )) 2 l N s =
( ( )) 1 r N s =
( ( )) ( ( )) 1 l N s r N s =
and

it follows from Step 10 that every admissible string


1
( 1) [ ( 18)] 1
l
sgn q

, = ,
0 1 2 3 4 5
{ } I i i i i i i = , , , , ,
must satisfy
0 1 2 3 4 5
{ 2 2 2 2 } ( 1) 6 i i i i i i + + = .
Hence the admissible strings are
1
2
3
4
5
{ 1 1 1 1 1 1}
{ 1 1 1 1 1 1}
{ 1 1 1 1 1 1}
{ 1 1 1 1 1 1}
{1 1 1 1 1 1}
I
I
I
I
I
= , , , , ,
= , , , , ,
= , , , , ,
= , , , , ,
= , , , , , .
From Step 11, for it follows that the stabilizing values
corresponding to must satisfy the string of inequalities:

1
I ( )
i d
k k ,
18
p
k =
2
1 0 0 2 0
2
1 1 1 2 1
2
1 2 2 2 2
2
1 3 3 2 3
2
1 4 4 2 4
2
1 5 5 2 5
( ) ( ) ( ) 0
( ) ( ) ( ) 0
( ) ( ) ( ) 0
( ) ( ) ( ) 0
( ) ( ) ( ) 0
( ) ( ) ( ) 0
i d
i d
i d
i d
i d
i d
p k k p
p k k p
p k k p
p k k p
p k k p
p k k p
e e e
e e e
e e e
e e e
e e e
e e e
+ <

+ <

+ <

+ >

+ <

+ >

Substituting for , , , , and in the above

expressions, we obtain
0
e
1
e
2
e
3
e
4
e
5
e
0
0 2699 4 6836
0 3666 10 0797
3 5358 3 912
13 5777 140 2055
i
i d
i d
i d
i d
k
k k
k k
k k
k k
<

. < .

. < .

. > .

. < .

The set of values of for which above holds can be

solved by linear programming and is denoted by . For ,
we have
( )
i d
k k ,
1
S
2
I
0
0 2699 4 6836
0 3666 10 0797
3 5358 3 912
13 5777 140 2055
i
i d
i d
i d
i d
k
k k
k k
k k
k k
<

. > .

. > .

. > .

. < .

The set of values of for which the above holds can also be
solved by linear programming and is denoted by .
( )
i d
k k ,
2
S
Similarly, we obtain
3 3
4 4
5 5
for
for
for
S I
S I
S I

= C
=C
=C
Then, the stabilizing set of values when is
given by
( )
i d
k k ,
18
p
k =
( 18) 1 2 5
1 2
x x
S S
S S
= , , ,
=
=
The set and the corresponding and are shown in
Fig. 2.
( 18)
S
1
S
2
S
Figure 2. The stabilizing set of values when . ( )
i d
k k , 18
p
k =
By sweeping over different values within the interval and
repeating the above procedure at each stage, we can generate
the set of stabilizing values. This set is shown in Fig.
3.
p
k ( 24 7513 1) . ,
( )
p i d
k k k , ,
Figure 3. The stabilizing set of values. ( )
p i d
k k k , ,
PID Controllers for Discrete-time Systems
Plant
( )
( )
( )
z
z
z
N z
G z
D z
=
where and are polynomials in the forward shift
operator .
The discrete-time PID controller is given by:
( )
z
N z ( )
z
D z
z
1 2
1 1
2
2
1 1 2
( )
1 1
( ) ( 2 )
z p i d
p i d p d d
z z
C z k k k
z z
k k k z k k z k
z z


+
= + +

+ + + +
=

1
1
w
z
w
+
=

1
1
( )
( )
( )
w
w
z z
N w
G z
D w
+

=
= ,
-domain PID controller

w
2
2( ) 2 4
( )
( ) 2 2
i p i p i d
k w k k w k k k
B w
A w w
+ + + + +
= .
+
-domain closed loop characteristic polynomial: w
2
( ) (2 2) ( ) ( 2( ) 2 4 ) ( )
p i d i p i p i d
w k k k w D w k w k k w k k k N w o , , , = + + + + + + +
and Hurwitz stability of this polynomial is equivalent to
stability of the original discrete time system. It is clear that we
can now proceed as in the previous section
Example Plant where

( )
( )
z
z
N z
D z
2
( ) 1
( ) 1 5 0 5
z
z
N z z
D z z z
= +
= . + . .
Using the bilinear transformation, we obtain the -domain plant
where

w
( )
( )
N w
D w
2
( ) 2 2
( ) 3
N w w w
D w w
=
= + .
Fig. 4 shows the stabilizing regions in the space of ( , , )
determined using the procedure outlined above.

p
k
d
k
i
k
Figure 4. The stabilizing region in the space of ( , , ).
p
k
i
k
d
k
PID Controllers for Continuous-time First
Order Systems with Time Delay
( ) G s We consider the feedback system of Figure 1 where the plant
is described by
( )
1
Ls
k
G s e
Ts

= .
+
represents the steady-state gain of the plant, the time delay,
and the time constant of the plant. The controller is of the PID
type,

k L
T
( )
i
p d
k
C s k k s
s
= + + .
The objective is to determine the set of controller parameters
( , , ). for which the closed-loop system is stable. A
complete solution to this problem has been obtained last year.
We provide a brief summary of these results.

p
k
i
k
d
k
[A] Open-loop Stable Plant
In this case . Furthermore, we make the standing
assumption that and .
0 T >
0 k > 0 L >
Theorem

The range of values for which a given open-loop stable plant,
with transfer function as in (??), continues to have closed
loop stability with a PID controller in the loop is given by

p
k
( ) G s
1 1 1
1 1
( ) ( )
p
T
k sin cos
k k L
o o o
(
< <
(

where is the solution of the equation
1
o
( )
T
tan
T L
o o =
+
in the interval . For values outside this range, there are no
stabilizing PID controllers. The complete stabilizing region is
given by: (see Fig. 5)

(0 ) t ,
p
k
1. For each the cross-section of the
stabilizing region in the space is the trapezoid T.

( )
i d
k k ,
1 1
,
p
k
k k
| |
e ,
|
\ .
2. For , the cross-section of the stabilizing region in the
space is the triangle .

1
p
k
k
=
( ) ,
i d
k k A
3. For each , the cross-
section of the stabilizing region in the space is
the quadrilateral Q.
| | ( )
1 1
1 1 1
, sin( ) cos( )
T
p u k k L
k k o o o e :=
( )
i d
k k ,
The parameters necessary for determining the
boundaries of T, and Q can be determined using

,
j j
m b , ,
j
w 1 2 j = ,
A
the following equations:
2
2 j
j
L
m
z
=
( ) ( )
sin cos
j j j j
j
L T
b z z z
kz L
(
= +
(

( ) ( ) ( )
sin cos 1
j
j j j j
z
T
w z z z
kL L
(
= + +
(

where are the real, positive solutions of

, 1, 2,...
j
z j =
( ) ( ) cos sin 0
p
T
kk z z z
L
+ =
arranged in ascending order of magnitude.

Figure 5. The stabilizing region of for:
i
k
d
k
( ) ,
i d
k k ( ) ( ) ( )
1 1 1 1
; ; .
p p p u
a k b k c k k
k k k k
< < = < <
[B] Open-Loop Unstable Plant In
this case Theorem

0 T <
A necessary and sufficient condition for the existence of a
stabilizing PID controller for the open-loop unstable plant (??)
is . If this condition is satisfied, then the range of
values for which a given open-loop unstable plant, with transfer
function can be stabilized using a PID controller is given
by

0 5
T
L
, ,> .
p
k
( ) G s
1 1 1
1 1
( ) ( )
p
T
sin cos k
k L k
o o o
(
< <
(

where is the solution of the equation

1
o
( )
T
tan
T L
o o =
+
in the interval . In the special case of , we have
For values outside this range, there are no stabilizing PID
controllers. Moreover, the complete stabilizing region is
characterized by: (see Fig. 6)

(0 ) t , 1
T
L
, ,=
1 2
.
t
o =
p
k
For each , the cross-
section of the stabilizing region in the space is the
quadrilateral Q.
| | ( )
1 1
1 1 1
sin( ) cos( )
T
p l k L k
k k o o o e := ,
( )
i d
k k ,
The parameters and , necessary for
determining the boundary of Q are as defined in the statement
of previous Theorem
j j
m b ,
j
w 1 2 j = ,
Figure 6. The stabilizing region of ( , ) for .

i
k
d
k
1
l p k
k k < <
PID Stabilization Algorithm for Time-Delay Plants:

1) Initialize and , where is the desired
number of grid points.
1
p k
k = ( )
1 1
1 u N k
step k
+
= + N
2) Increase as follows: .

p
k
p p
k k step = +
3) If then go to Step 4. Else, terminate the algorithm.
p u
k k <
4) Find the roots and 1
z
2
z
5) Compute the parameters and , associated with
the previously found by using (1) and (1).

j
m
j
b 1 2 j = ,
j
z
6) Determine the stabilizing region in the - space using
Fig. 5.

i
k
d
k
7) Go to Step 2.

Example

Consider the PID stabilization problem for a plant described by the differential
equation
( )
0 5 ( ) 0 5 ( 4)
dy t
y t u t
dt
= . + . .
- This process can also be described by the transfer function with the
following parameters: , sec, and sec. We use the Theorem above
to find the range of values for which a solution to the PID stabilization problem
exists.

( ) G s
1 k = 2 T = 4 L =
p
k
- We first compute the parameter satisfying the following equation
1
(0 ) o t e ,
tan( ) 0 3333 o o = . .
- Solving this equation we obtain . Thus, from (??) the range of
values is given by

1
2 4557 o = .
p
k
1 1 5515
p
k < < . .
- We now sweep over the above range of values and determine the stabilizing
set of values at each stage using the previous algorithm. These regions are
sketched in Fig. 7.

p
k
( )
i d
k k ,
- Any PID gains selected from these regions will result in closed-loop stability and
any gains outside will result in instability.
Now, consider the following performance specifications:

1. Settling time secs;

60 s
Figure 7. The stabilizing region of ( , , ) values for the PID controller in
Example ??.
p
k
i
k
d
k
2. Overshoot . 20% s
We can obtain the transient responses of the closed-loop system for the
( , , ) values inside the regions depicted in the Fig. In general we
also need some tolerance around the controller parameters, that is we want
the controller to be controller-robust or non-fragile. Thus we only consider
PID gains lying inside the following box defined in the parameter space:

p
k
i
k
d
k
0 1 1 0 1 0 3 and 0 5 1 5
p i d
k k k . s s , . s s . . s s . .
By searching over this box, several ( , , ) values are found to meet the desired
performance specifications. We arbitrarily set the controller parameters to: ,
, . Fig. below shows the step response of the resulting closed-loop
p
k
i
k
d
k
0 3444
p
k = .
0 1667
i
k = . 0 8333
d
k = .
system. It is clear from the figure that the closed-loop system is stable, the output
tracks the step input signal and the performance specifications are met.
( ) y t
Figure 8. Time response of the closed-loop system for Example
The figure also shows the responses of the closed-loop systems for the case of a
PID controller designed using the Cohen-Coon method
and the Ziegler-Nichols method Notice that in these
cases also the system is stable and achieves setpoint following. However, the
responses are much more oscillatory.
( 0 9180 0 1456 0 9845)
p i d
k k k = . , = . , = .
( 0 6 0 075 1 2)
p i d
k k k = . , = . , = .
PID Controller Design for Performance Specs
- In many situations control system performance can be specified
by a frequency domain inequality or equivalently an norm constraint
on a closed loop transfer function :

H

( )
( )
( )
N s
D s
G s =
( ) G s

< .
- It has been shown by us that the above condition is equivalent to Hurwitz
stability of the complex polynomial family:

( ) ( ) [0 2 ]
j
D s e N s
u
u t + , e , .
- In our PID design problem the polynomials will have the PID
gains embedded in them and the set of parameters achieving specifications is
given by those achieving simultaneously the stabilization of the complex
polynomial family as well as the real closed loop characteristic polynomial. It
turns out that the set of PID gains achieving stabilization of a complex polynomial
family and therefore attaining the specifications can be found by an extension of
the algorithm given for the real case.
( ) ( ) D s N s ,
2
( ) ( ) ( ) ( )
p i d d p i
c s k k k L s k s k s k M s , , , = + + +
-
where and are given complex polynomials. The results on PID
stabilization presented earlier have been extended to this complex stabilization
problem (details omitted)

( ) L s ( ) M s
Synthesis of PID Controllers
Consider the following closed-loop transfer functions: considered:
- The sensitivity function:


- The complementary sensitivity function:


- The input sensitivity function:



- Various performance and robustness specifications can be captured by using the norm of weighted
versions of the transfer functions. When is a PID controller, the transfer functions (?)-(?) can all be
represented in the following general form:




,
) ( ) ( 1
1
) (
s G s C
s S
+
=
,
) ( ) ( 1
) ( ) (
) (
s G s C
s G s C
s T
+
=
,
) ( ) ( 1
) ( ) (
) (
s G s C
s G s C
s T
+
=
, (s
Tcl
H
) ( ) ( ) (
) ( ) ( ) (
) , , , (
2
2
s N k s k s k s sD
s B k s k s k s A
k k K s T
i p d
i p d
d i p
+ + +
+ + +
=
H
) (s C
For the transfer function and a given number , the standard
performance specification usually takes the form:


where is a stable frequency-dependent weighting function that is selected to capture the desired design
objectives at hand.

Define the polynomials and as follows:

) , , , (
d i p
k k K s T
H
< || ) , , , ( ) ( ||
d i p cl
k k k s T s W
0 >
( ) W s
( )
p i d
s k k k o , , ,
( )
p i d
s k k k | u , , , , ,
2
( ) ( ) ( ) ( )
p i d i p d
s k k k sD s k k s k s N s o
A
, , , + + +
=
and
2
1
( ) [ ( ) ( ) ( ) ( )]
1
( )[ ( ) ( ) ( ) ( )]
j
p i d d n
j
d p i d n
s k k k sW s D s e W s A s
k s k s k W s N s e W s B s
u
u
| u

A
, , , , , +
+ + + + .
=
The performance problem reduces to the simultaneous satisfaction of the following conditions:

1. is Hurwitz;
2. is Hurwitz for all in ;
3.

H

( )
p i d
s k k k o , , ,
( )
p i d
s k k k | u , , , , , u
[0 2 ) t ,
( ) ( )
cl p i d
W T k k k , , , , ,<
The above equivalence can be used to determine stabilizing values such that
the -norm of a certain closed-loop transfer function is less than a prescribed level. This
is illustrated using the following example.

( )
p i d
k k k , ,
H

Example

Consider the plant where


( )
( )
( )
N s
D s
G s =
2
( ) 1
( ) 0 8 0 2
N s s
D s s s
=
= + . .
We consider the problem of determining all stabilizing PID gain values for which
where is the complementary
sensitivity function:
1 || ) , , , ( ) ( || <
d i p
k k k s T s W
( )
p i d
T s k k k , , ,
2
2 2
( )( 1)
( )
( 0 8 0 2) ( )( 1)
d p i
p i d
d p i
k s k s k s
T s k k k
s s s k s k s k s
+ +
, , , =
+ . . + + +
and
0 1
( )
1
s
W s
s
+ .
= .
+
We have to satisfy:
1. is Hurwitz;
2. is Hurwitz for all
in ;
3.

These sets are shown in the following:
2 2
( ) ( 0 8 0 2) ( )( 1)
p i d d p i
s k k k s s s k s k s k s o , , , = + . . + + +
2 2
( 1 ) ( 1)( 0 8 0 2) ( )[( 1)( 1) ( 0 1)( 1)]
j
p i d d p i
s k k k s s s s k s k s k s s e s s
u
| u , , , , , = + + . . + + + + + + .
1
( ) ( ) 1
d
d
k
p i d k
W T k k k
+
, , , , ,=, ,<
u
[0 2 ) t ,
Figure 9: The set
(1 0 35)
S
, .
(2 0 35) [0 2 ) (2 0 35 )
S S
u t u , . e , , . ,
=
Figure 10: The set
Figure 11: The set of stabilizing values for
which
( )
p i d
k k k , ,
1 || ) , , , ( ) ( || <
d i p
k k k s T s W
Example: Robust Performance
Consider the plant where
( )
( )
( )
N s
D s
G s =
2
( ) 15
( ) 1
N s s
D s s s
=
= + .
Then the sensitivity function and complementary sensitivity function are:
2
2 2
2
2 2
( 1)
( )
( 1) ( )( 15)
( )( 15)
( )
( 1) ( )( 15)
p i d
d p i
d p i
p i d
d p i
s s s
S s k k k
s s s k s k s k s
k s k s k s
T s k k k
s s s k s k s k s
+
, , , = ,
+ + + +
+ +
, , , = .
+ + + +
The weighting functions are chosen as: and . We know
that stabilizing values meeting the robust performance specification
exist if and only if the following conditions hold:
0 2
1 0 2
( )
s
W s
.
+ .
=
0 1
2 1
( )
s
s
W s
+ .
+
=
( )
p i d
k k k , ,
Example: Robust Performance Cont.
1. is Hurwitz;
2.
is Hurwitz for all
and for all ;
3.
The procedure for determining the set of values satisfying conditions (1), (2) and
(3) is similar to that presented in the previous example.
2 2
( ) ( 1) ( )( 15)
p i d d p i
s k k k s s s k s k s k s o , , , = + + + +
1 2 1
( ) ( ) ( ) ( ) 1
d
d
k
p i d p i d k
W S k k k W T k k k
+
, , , , , +, , , , ,=, ,<
( )
p i d
k k k , ,
+ + + + + + + = ) 1 )( 1 )( 2 . 0 ( ) 1 )( 1 )( 2 . 0 ( ) , , , , , (
2 2
s s s s e s s s s s k k k s
j
d i p
u
| u
)] 15 )( 1 . 0 )( 2 . 0 ( ) 15 )( 1 )( 2 . 0 )[ (
2
+ + + + + + + s s s e s s s k s k s k
j
i p d
|
Figure 12: The set of values for which ( )
p i d
k k k , ,
1 2
( ) ( ) ( ) ( ) 1
p i d p i d
W s S s k k k W s T s k k k

, , , , , + , , , , , <
PID Controller Design with Guaranteed Gain and Phase Margins
We consider the problem of designing PID controllers that achieve pre-specified gain and phase margins for a
given plant. Let and denote the desired upper gain and phase margins respectively. From the
definitions of the upper gain and phase margins, it follows that the PID gain values achieving
gain margin and phase margin must satisfy the following conditions:
m
A
m
u
( )
p i d
k k k , ,
m
A
m
u

1. is Hurwitz for all and
2. is Hurwitz for all
2
( ) ( ) ( )
d p i
sD s A k s k s k N s + + +
[1 ]
m
A A e , ;
2
( ) ( ) ( )
j
d p i
sD s e k s k s k N s
u
+ + +
[0 ]
m
u u e , .
Thus the problem to be solved is reduced to the problem of simultaneous stabilization of two
families of polynomials.
Example
Consider the plant where
( )
( )
( )
N s
D s
G s =
4 3 2
( ) 2 1
( ) 3 4 7 9
N s s
D s s s s s
=
= + + + + .
In this example, we consider the problem of determining all gain values that
provide a gain margin and a phase margin . A given set of
values will meet these specifications if and only if the following conditions hold:
( )
p i d
k k k , ,
3 0
m
A > .
40
m
u >
( )
p i d
k k k , ,
1. is Hurwitz for all
;
2. is Hurwitz for all
.
4 3 2 2
( 3 4 7 9) ( )(2 1)
d p i
s s s s s A k s k s k s + + + + + + +
[1 3 0] Ae , .
4 3 2 2
( 3 4 7 9) ( )(2 1)
j
d p i
s s s s s e k s k s k s
u
+ + + + + + +
[0 40 ] u e ,
Again, the procedure for determining the set of values is similar to that presented
before. The resulting set is sketched in Fig. 13.
( )
p i d
k k k , ,
Figure 13: The set of values for which the resulting closed loop system
achieves a gain margin and a phase margin .
( )
p i d
k k k , ,
3 0
m
A > . 40
m
u >
Concluding Remarks
- Similar results have been obtained for first order (lead/lag) controllers
- Extensions to arbitrary fixed order controllers under study
- Software development planned
- Time domain specs. to be incorporated
- Extension to MIMO systems
- Applications

References
- Astrom K. J. strm and T. Hgglund, PID Controllers: Theory, Design, and Tuning,
Instrument Society of America, North Carolina, 1995.
- A. Datta, M. T. Ho and S. P. Bhattacharyya, Structure and Synthesis of PID Controllers,
Springer-Verlag, 2000.
- H. Xu, A. Datta and S. P. Bhattacharyya, Computation of All Stabilizing PID Gains for
Digital Control Systems, IEEE Transactions on Automatic Control, Vol. AC-46, No. 4, 647-
652, April 2001.
- G. J. Silva, A. Datta and S. P. Bhattacharyya, New Results on the Synthesis of PID
Controllers, IEEE Transactions on Automatic Control, Vol. 47, No. 2, 241-252, February 2002.
L.H.Keel, J.I.Rego and S.P.Bhattacharyya, A New Approach to Digital PID Controller
Design" IEEE Trans. Aut. Contr. Vol. AC-48(4), pp.687-692, April 2003.
Motivation
Design of low order controllers
Whether there exists a stabilizing controller of a given
order
If so, what is the complete set of stabilizing controllers
of a given order
In terms of a given performance metric, is there a
stabilizing controller that achieves a desired
performance?
If there are several performance metrics, what is the set
of stabilizing controllers that meet performance
objectives in terms of these metrics.
Motivation
Design of decentralized controllers for a collection of
vehicles with a given information structure:

What are the set of stabilizing controllers that achieve a
certain performance (e.g. spacing error attenuation)?

Such requirements are important for ``scalability of
stability for arbitrarily large collections

Problem Statement
Given Data (polynomials):

Questions:
Do there exist controller gains
such that the following polynomial is Hurwitz?


If so, what is the set of all such controller gains that
make the above polynomial Hurwitz?
0 1
( ), ( ), , ( )
l
P s P s P s
1 2
( , , , )
l
K k k k =
0 1 1 2 2
( , ) : ( ) ( ) ( ) ( )
l l
P s K P s k P s k P s k P s = + + + +
Example
First Order controller for a MISO system

Plant:

Controller:

Characteristic polynomial:



1
( )
( ) ( )
( )
r
i
i
i
p
N s
Y s U s
D s
=
=

( ) ( )
i i
i
a s b
U s Y s
s c
+
=
+
3 2 1
2 2 1
1 1 1 1
( ) ( )
( ) ( ) ( )
( )
( , ) ( ) ( ) ( ) ( ) ( ) ( )
r
r
o
p p r r r r
P s P s
P s P s P s
P s
P s K sD s c D s a sN s b N s a sN s b N s
+
= + + + + + +
Construction of the set of stabilizing controllers
Hermite-Biehler (HB) Theorem: Let





P(s,K) is Hurwitz iff
The constant coefficients of even and odd polynomials are of the
same sign
Roots of the even and odd polynomial are real and interlace
2
2
( , ) Re( ( , ))
1
( , ) Im( ( , ))
e
o
P w K P jw K
P w K P jw K
w
=
=
Idea of the proof of Hermite-Biehler Theorem






1 2
( , ) ( )( ) ( )
n
P jw K jw s jw s jw s Z = Z
1 2
( ) ( ) ( )
n
jw s jw s jw s = Z +Z + +Z
( )
e o
P jwP = Z +
Construction of stabilizing sets
There exists a stabilizing K (meaning P(s,K) is Hurwitz) iff
All coefficients of P(s,K) are of the same sign
there exist (n-1) frequencies
such that






0 1 1
0
n
w w w

= < < <
2 2
1
1 2 2
1
( 1) ( , ) ( , ) 0,
( 1) ( , ) ( , ) 0,
l
e l e l
l
o l o l
P w K P w K
P w K P w K
+
+
+
<
<
0,1, , 2 l n =
How many LPs for a set of frequencies?
Given a set of (n-1) frequencies
Only the following two LPs must be checked:





0 1 1
0
n
w w w

= < < <
2 2
1 1
2 2
2 2
2 2
3 3
2 2
4 4
(0, ) 0, (0, ) 0
( , ) 0, ( , ) 0
( , ) 0, ( , ) 0
( , ) 0, ( , ) 0
( , ) 0, ( , ) 0
e o
e o
e o
e o
e o
P K P K
P w K P w K
P w K P w K
P w K P w K
P w K P w K
> >
< >
< <
> <
> >
2 2
1 1
2 2
2 2
2 2
3 3
2 2
4 4
(0, ) 0, (0, ) 0
( , ) 0, ( , ) 0
( , ) 0, ( , ) 0
( , ) 0, ( , ) 0
( , ) 0, ( , ) 0
e o
e o
e o
e o
e o
P K P K
P w K P w K
P w K P w K
P w K P w K
P w K P w K
< <
> <
> >
< >
< <
What do the feasible sets of LPs mean?
Given a set of (n-1) frequencies,
The two LPs associated with these frequencies indicate all
controllers, K, which stabilize the plant and
which place roots of and alternatively in the
disjoint intervals


Which ensure that all coefficients of P(s,K) are of the
same sign







0 1 1
0
n
w w w

= < < <
e
P
o
P
1 1 2 2 3 2 1
(0, ), ( , ), ( , ), , ( , )
n n
w w w w w w w

Construction of the set of all stabilizing controllers
One can identify, in a unique way, any feasible LP with
A set of (n-1) increasing frequencies
a binary number which indicates the sign of the
coefficients

Problem of determining the set of controllers can therefore
be reduced to the search for all these n real numbers
Store all such sets of n numbers for purposes of design
Compactify (0, ) (0, 1) using




2
, (0,1)
1
u
w u
u
= e

Example (Anderson, IEEE TAC 1975)

Plant:





Controller:




0 1 0 0
0 0 1 0
0 13 0 1
0 5 1
1 1 0
x x u
y x
( (
( (
= +
( (
( (

(
=
(


| |
1 2
u k k y =
Example
Characteristic polynomial:


Necessary conditions:









3 2
1 2 1 2 1 2
( , , ) ( 5 13) P s k k s k s k k s k = + + +
1
2 1
2
0
( 5 13) 0
0
k
k k
k
>
>
>
Example
Even and odd polynomials:












3 2
1 2 1 2 1 2
2 2
1 2 1 2
2 2
1 2 2 1
( , , ) ( 5 13)
( , , )
( , , ) ( 5 13)
e
o
P jw k k jw k w k k jw k
P w k k k w k
P w k k w k k
= + +
= +
= +
Example
There must exist two frequencies: so
that








1 2
0 w w < <
1 2 2
1 2 2 1
2 2
1 1 2 1 1 2
2 2
1 1 2 1 2 1
2 2
2 1 2 1 2 2
2 2
2 1 2 2 2 1
(0, , ) 0;
(0, , ) 5 13 0;
( , , ) 0;
( , , ) 5 13 0;
( , , ) 0;
( , , ) 5 13 0;
e
o
e
o
e
o
P k k k
P k k k k
P w k k k w k
P w k k w k k
P w k k k w k
P w k k w k k
= >
= >
= + <
= + >
= + <
= + <
Example
There should exist such that







1 2
0 1 u u < < <
1 2 2
1 2 2 1
1 1 2 1 1 2 1
1 1 2 1 2 1 1
2 1 2 1 2 2 2
2 1 2 2 2 1 2
(0, , ) 0;
(0, , ) 5 13 0;
( , , ) (1 ) 0;
( , , ) ( 5 13)(1 ) 0;
( , , ) (1 ) 0;
( , , ) ( 5 13)(1 ) 0;
e
o
e
o
e
o
R k k k
R k k k k
R u k k k u k u
R u k k u k k u
R u k k k u k u
R u k k u k k u
= >
= >
= + <
= + >
= + <
= + <
Example - Results
Partition (0,1) and search over the partition







Example - Results
Partition (0,1) and search over the partition







Example - Results
Partition







Relevant results from recent literature
Hermite-Biehler theorem is used to
Synthesize PID controllers (by Bhattacharyya, Keel,
Datta, Ho etc)
obtain a parametrization of all Hurwitz polynomials of a
given degree by Djaferis et.al (May 2003, IEEE TAC).
It maps the interior of a non-negative monotone cone of n
frequencies into the set of all monic Hurwitz polynomials of
degree n in a bijective manner

Purpose of Outer Approximation

Theorem: If the set of strictly proper controllers of order r
is bounded (but not empty), then r is the minimal order
of stabilization

Boundedness of an outer approximation of the set of
stabilizing controllers of a given order ensures that one
is working with minimal order of stabilization
Outer Approximation
Descartes Rule of signs:
The number of sign changes in the coefficients of a
polynomial are greater than or equal to the number of its
real, positive roots and the difference is always even.
If all roots of a polynomial are real, then the number of
sign changes in the coefficients is exactly equal to the
number of real, positive roots of the polynomial

Application of Descartes rule to P(s,K) is equivalent to
requiring all its coefficients be of the same sign (leads to
two LPs)




Outer Approximation
A Generalization of Descartes rule (due to Poincare):
(See Polya and Szego) Let R(s) be a polynomial.

1. The number of sign changes in the coefficients of
monotonically decrease with k

2. As k approaches infinity, the number of sign changes in
the coefficients of is exactly equal to the
number of real positive roots of R(s)
( 1) ( )
k
s R s +
( 1) ( )
k
s R s +
Sketch of the methodology
Applying the Generalization of Descartes rule (due to
Poincare): It is necessary for P(s,K) to be Hurwitz that
1. For every + integer k, the number of sign changes of
is the same as the degree of the even
polynomial
2. Similar statement for the odd polynomial

For every k, these two conditions generate a number of LPs;
the union of the feasible sets of LPs contains the set of
stabilizing controllers (outer approximation)
2 2
( 1) ( , )
k
e
w P w K +
2
( , )
e
P w K

Idea of the proof of Poincares result
For k sufficiently large, the number of sign changes in the
coefficients of

is the same as the number of sign changes of the sequence


where

( 1) ( )
k
s R s +
( ) (1 ) ( )
1
n
u
R u u R
u
=

1 2 1
{ ( ), ( ), , ( )}
k
R R R
k k k

Outer Approximation - Interlacing
For two polynomials to have all real positive roots that
interlace, given any real positive number, the difference
in the number of roots of the two polynomials to the
left of the given number is either 0 or 1 or -1.

Using this fact, one can construct an outer approximation
which has the following feature:
If K is destabilizing, there is an iteration (akin to the
exponent k in the Poincares result) such that the kth
and subsequent outer approximants will not include K.

Problems that can be tackled
Simultaneous stabilization of a discrete number of
plants

Synthesis of controllers with a specified gain margin
Can be posed as a problem of stabilizing a one-parameter
family of plants, where the parameter belongs to a compact set,
which can be discretized and converted to the problem above


Relevant results from recent literature
Henrion, Sebek and Kucera: Idea of their scheme:
They use the result that a polynomial P(s) is Hurwitz iff
there exists a Hurwitz polynomial Q(s) of the same
degree such that P(s)/Q(s) is SPR.
Q(s) central polynomial guessed.
Requiring SPR of P(s)/Q(s) is equivalent to requiring
another polynomial, with coefficients affine in
controller parameters, to be non-negative (Siljaks result
from 1973)
They use Sum-of-Squares method and solve using LMIs

Relevant results from recent literature
Descartes Rule of signs Underutilized tool

Using the Poincares generalization, one can show the
following result:

If a plant P(s) does not have real, non-minimum phase
zeros, there is a two parameter stabilizing compensator
that achieves a non-negative closed loop impulse
response (only the order can be quite large)



Conclusions
1. Presented a method to construct the set of all
stabilizing controllers of fixed order
2. Presented a method to get a bound for the set of all
stabilizing controllers of fixed order
3. Both approximations exploit interlacing property of
Hurwitz polynomials as described by the Hermite-
Biehler Theorem
4. The proposed methodology has limitations
it does not address the question of existence of
stabilizing controllers in a crisp manner

Work In Progress
Synthesis of the set of controllers that acheive a
specified performance
Synthesis of fixed structure controllers for MIMO
systems
Synthesis of controllers which are based on the
empirical response data of the plant
Experiments