Robust Stability and Performance: ME8281 - Last Updated: April 21, 2008 (C) Perry Li

ME8281 - Last updated: April 21, 2008 (c) Perry Li
Robust stability and Performance
Topics: ([ author ] is supplementary source)
Sensitivities and internal stability (Goodwin 5.1-5.4)
Modeling Error and Model Uncertainty (Goodwin

4.12-4.13, Doyle 4.1)
Robust stability (Goodwin 5.7-5.10, Doyle 4.2)
Robust performance (Doyle 4.3-4.4, Goodwin 5.9)
Loop-shaping technique (Doyle 7.1) [We will cover

this only briefly]
Innovation feedback and affine parameterization

approach to tuning S and T (see Goodwin Chapters
15, 16, 18.5-6 for details)
The following topic(s) are for your information:
Performance limitation (Glad 7.3-7.4, Doyle 6.1-6.2,

Goodwin [8.6],9.1-9.3)
M.E. University of Minnesota 137

Nominal Sensitivity Functions

Y (s) Y (s) Go(s)C(s)
To(s) = = = (18)
Dm(s) R(s) 1 + Go(s)C(s)

Y (s) 1
So(s) = = (19)
Do(s) 1 + Go(s)C(s)

Y (s) Go(s)
Sio(s) = = (20)
Di(s) 1 + Go(s)C(s)

U (s) C(s)
Suo(s) = = (21)
Do(s) 1 + Go(s)C(s)
To - Complementary sensitivity (goal: small for

noise, 1 for command following)
So - Sensitivity (goal: small)
Sio - Input disturbance sensitivity (goal: small)
Suo - Control sensitivity (goal: small)

Internal Stability
The nominal loop is internally stable if and only if all

eight transfer functions below is stable:

Go (s)C(s) Go (s) 1 Go (s)C(s) H(s)R(s)
C(s) Go (s)C(s) C(s) C(s)
Yo (s) Di (s)
Uo (s)
=
Do (s)
1 + Go(s)C(s)
Dm (s)
Let C(s) = P (s)/L(s), and Go(s) = Bo(s)/Ao(s).
Proposition The system is internally stable if and only

if the roots of the nominal closed loop characteristic
equation
Ao(s)L(s) + Bo(s)P (s) = 0
all lie in the open left half plane.

Example Consider
s s + 2
Go(s) = , C(s) =
(s + 4)(s + 2) s
where pole-zero cancellations at s = 2 and s = 0

occur.
Complementary sensitivity is stable:
G(s)C(s) 1
To(s) = Y (s)/Dm(s) = =
1 + Go(s)C(s) s + 5
Sensitivity is also stable:
s+4
S(s) = 1 To(s) =
s+5
However, control sensitivity is marginally stable:
U (s) C(s) (s + 2)(s + 4)

Suo(s) = = =
Dm(s) 1 + Go(s)C(s) (s + 5)s
And, input disturbance sensitivity is unstable:
Y (s) s
Sio(s) = =
Di(s) (s + 2)(s + 5)

The effect of unstable pole at s = 2 shows up in

Sio as input disturbance will drive the output Y to
be unbounded. This effect is not observed by the
controller as it is blocked by its zero.
The effect of unstable pole at s = 0 shows up in

the control sensitivity so that output disturbance or
measurement noise will drive the control output to
be unbounded. This effect is not observed at the
output as it is blocked by the plants zero.
Characteristic equation has unstable and marginally

stable poles:
(s + 2)s + (s + 4)(s + 2)s = (s + 2)(s + 5)s
predicting correctly that the system is NOT

internally stable.

Modeling Error
For linear systems, if Go(s) is the nominal system, and

the actual system is G(s), then define:
Additive uncertainty:
G(s) = G(s) Go(s)
Multiplicative uncertainty:
Go(s)G(s) = G(s) Go(s)
or
G(s) Go(s)
G(s) =
Go(s)

Example: Time delays -
The transfer function of time delay is e s. Since it is

not rational, one often models it as
k
s s + 2k
e
s + 2k
where k = 0, 1, 2, . . ..
The additive modeling error is:
k
s s + 2k
G(s) = e
s + 2k
The multiplicative modeling error is:
" k # k
s + 2k s + 2k
G(s) = e s / .
s + 2k s + 2k

Multiplicative Modeling error of approximation of various orders for 1s time delay

2
10
0
10
2
10
MME
4
10
6 k=0
10 k=1
k=2
8
k=3
10 k=5
10
10
2 1 0 1 2
10 10 10 10 10
rad/s
Other examples:
Uncertain pole location:
1/(s + Am) 1/(s + A) where A [A0, A1]
Neglected (possibly structural) dynamics:
n2 /(s2 + 2ns + n2 ) 1.
Neglected compressibility effect, etc...

Robust Stability
Let Go(s) be the nominal plant. Consider a family of

plants characterized by:
P := {G(s) = (1 + Wu(s)(s))Go(s)}
where
G(s) = Wu(s)(s) is the multiplicative model

uncertainty
Wu(s) is a given stable uncertainty weighting

function
(s) is the uncertainty itself, which is any stable

transfer function with |(j)| < 1 for all .
The question of robust stability is whether a controller

designed for Go(s) also stabilizes the any plant G(s)
P . If so, we say that the controller provides robust
stability.

Robust Stability Theorem: Assume that the

controller C(s) internally stabilizes the nominal plant
Go(s). Suppose that G(s)C(s) and Go(s)C(s) have
the same number of unstable (openloop) poles on the
open right half plane. Then,
1. If |G(j)| |To(j)| < 1 for all (in other

words, kG Tok < 1) then the controller C(s)
internally stabilizes the perturbed plant G(s).
2. C(s) provides robust stability for the plant set P if

and only if
kWu Tok < 1 (22)
where To(s) is the complementary sensitivity
function:
Lo(s)
To(s) =
1 + Lo(s)
where Lo(s) = Go(s)C(s).
Remark:
kF k := sup|F (j)| is the so called infinity
norm of the transfer function F (s).
Every plant G(s) P corresponds to a G(s) that

satisfies kG Tok < 1. So the extra interesting
aspect of robust stability condition is the necessity.

Before we give a proof of this theorem, recall the

Nyquist theorem:
Nyquist Theorem: Suppose that L(s) has P unstable

poles on the open RHP. Then, the Nyquist plot (i.e.
the plot of L(s) on the complex plane, as s traverses
the Nyquist contour (i.e. the imaginary axis, indented
to the right in case of poles of L(s) on the imaginary
axis, and a half circle at infinity encircling the RHP),
encircles the (1, 0) point clockwise N = Z P times,
where Z is the number of unstable poles of the closed
loop system:
L(s)
1 + L(s)
Corollary: The closed loop system is stable if and only

if the Nyquist plot of L(s) encircles the (1, 0) point
P times in the counter-clockwise direction.
This is obtained by setting Z = 0.
Note: P is the number of open-loop unstable poles.

Return now to the robust stability theorem.
Notice that the perturbed plant lies in a family of

disks of size |Lo(j)Wu(j)| centered at Lo(j). The
perturbed plant is at a distance |G(j)Lo(j)| from
Lo(j).
Geometric interpretation: Since
|ToWu|s=j = |Lo(j)Wu(j)|/|1 + Lo(j)|
and 1 + Lo(j) is the distance between (1, 0) and

Lo(j), |ToWu|s=j < 1 if and only
|Lo(j)Wu(j)| < |1 + Lo(j)|
Thus, the robust stability theorem states that the

family of disks of size |Lo(j)Wu(j)| centered at
Lo(j) should not contain the (1, 0) point.

Proof of Robust Stability:

Item 1 and sufficiency in item 2:
Show that the number of encirclement of (-1, 0)

does not change
The distance from Lo(j) to (1, 0) is:
|1 + Lo(j)|
However, since

Lo(j)
|To(j)G(j)| < 1
G(j) < 1
1 + Lo(j)
This implies that
|G(s)Lo(j)| < |1 + Lo(j)| = distance to (-1, 0)
Hence, the perturbed plant cannot reach (and hence

change the encirclement of) the (1, 0) point.
This is the case because if the perturbed plant does

change the encirclement of (1, 0) there would
be a < 1 such that the perturbed plant (1 +
G(s))Lo(s) touches (1, 0) at some s = j1.

Necessity in item 2:
Suppose that kWu Tok 1 (i.e. the robust

stability condition not satisfied) hence,
|Wu(j1) To(j1)| = 1 1
at some 1.
We need to construct a (s) with kk < 1 such

that, with G(s) = Wu(s)(s), (1+G(j))Lo(s)
touches the (1, 0) point.
To do this consider (s) of the form:
1 (s )
(s) =
(s + )
Note that |(j)| = 1/|| for all (i.e. all pass

filter) so that:
1
(j) = ej(,)

where (, ) = 2tan1(/). Thus, for any

, by choice of , it is possible to achieve any
(, ) (0, ].

Thus, we can choose = 1 and such that
1
1+Lo(j1) = Wu(j1)ej(,1) = Lo(j1)G(j1)
1
This makes the perturbed Nyquist plot touch the

(-1, 0) point.
QED.
Remarks:
The perturbed closed loop system can be formulated
into a closed loop system between G1(s) =
G(s) = Wu(s)(s) and G2(s) = To(s).
Sufficiency part of robust stability is a special case

of the Small Gain Theorem (SGT).

Small Gain Theorem (SGT): Let G1 and G2 be

(possibly nonlinear) stable systems with finite input-
output gains. Let kG1k and kG2k denote their
respective gains, i.e. their induced norms. If
kG1k kG2k < 1,
then the closed loop system consisting of G1 and G2

will also be stable.
Remarks: If we consider Gi : u() 7 y() then, the
induced norm (gain) of Gi is defined to be:
ky()k
kGiki := supu()
ku()k
where ku()k and ky()k are the respective signal norms

of the input and output. By using different signal
norms, different induced norms of the system can be
obtained.
For linear systems, it turns out that kGk is the
induced 2-norm. i.e. the input and output are
measured using the 2-norm:
Z 12
ku()k2 = u(t)2dt .


Therefore, the sufficiency of (22) is exactly what

is provided by the Small Gain Theorem. What is
interesting for LTI systems is that (22) is also necessary
condition.

Robust Performance
We assume that the performance is specified by the
smallness of the achieved sensitivity S(s).
1 1
S(s) = =
1 + G(s)C(s) 1 + L(s)
Nominal performance:
1
kWpSok < 1; So(s) =
1 + Lo(s)
where Lo(s) = Go(s)C(s), Wp(s) is the sensitivity

performance weight.

How to specify sensitivity weighting Wp ?
For example, output disturbance is
do = k + asin(t + )
where
DC disturbance: k [0, 10],
AC disturbance: [2, 4]rads1, a A.
If we would like the effect of disturbance to be smaller

than 1, then, we would choose |S(j0)| < 1/10 to
satisfy DC disturbance; and |S(j)| < 1/A for
[2, 4]rads1 to satisfy AC disturbance requirements.
Similar methodology can be used to specify

requirements for Suo, Sio etc.

Consider a family of plants characterized by:
P := {G(s) = (1 + Wu(s)(s))Go(s)}
G(s) = Wu(s)(s) is the multiplicative model

error
Wu(s) is a given stable uncertainty weighting

function
(s) is the uncertainty itself, which is any stable

transfer function with |(j)| < 1 for all .
Perturbed sensitivity:
1 So
S(s) = =
1 + (1 + G)Lo(s) 1 + G(s)To(s)
Lo (s)
where To(s) = 1+Lo (s) .
We are interested in the controller C(s) stabilizing all

plants in set P and also satisfying the performance
specification kWpSk < 1.
Robust performance means: kk < 1,

Wp So
kWuTok < 1 and 1 + WuTo < 1 (23)


Theorem: A necessary and sufficient condition for

robust performance is:
k|WpSo| + |WuTo|k < 1 (24)
Sketch of Proof:
Sufficency Suppose that (24) is satisfied.
Clearly (23) implies that kWuTok < 1

WpSo
We need only show that 1+WuTo < 1 for any

allowable (s).
Since |WpSo| + |WuTo| < 1, we have

Wp So

1 |WuTo| <1

It is easy to see that for any k(s)k < 1,

Wp So Wp So

1 |WuTo|

1 + |WuTo|


(worst case with kk < 1 is for RHS to equal

LHS) thus, we have

Wp So

1 + WuTo <1

Necessity: Assume that (23) is true.
Construct a such that
|WpSo| |WpSo|
=
1 |WuTo| 1 + WuTo
at the frequency where the LHS is maximized. Since

the RHS < 1 by robust performance,

Wp So

1 |WuTo| <1

and hence (24) is true.

Geometric interpretation:
The robust performance theorem states that at all
, disks of size |Lo(j)Wu(j)| centered at Lo(j)
should not intersect disk with radius |Wp(j)| the
(1, 0) point.
This is because:
|Wp(j)So(j)| = |Wp(j)|/|1 + Lo(j)|

|Wu(j)To(j)| = |Wu(j)Lo(j)|/|1 + Lo(j)|
Thus, k|WpSo| + |WuTo|k < 1 if and only if:
|Wp(j)| + |Wu(j)Lo(j)| < |1 + Lo(j)|
the latter is the distance between (1, 0) and Lo(j).

|Wp(jw)|
(1, 0)
|Wu(jw)| |Lo(jw)|
Lo(jw)

Robust performance can be achieved by robust stability

and nominal performance
Proposition: If a controller provides
Nominal performance kWpSok < 1 for a

performance weighting Wp(s).
Robust stability kWuTok < 1 for an uncertainty

weighting Wu(s).
then the controller provides for robust performance

w.r.t
1 1
k| WpSo| + | WuTo|k < 1.
2 2
i.e. when the performance requirement, and
uncertainty requirements are halved.
Proof:
1 1 1 1
k | W p S o | + | W u T o | k < k + k 1 Q.E.D.
|2 {z } |2 {z } 2 2
<0.5 <0.5
We can also run the argument backwards. To solve

the robust performance criteria for:
k|Wp So| + |Wu To|k < 1.

we can design a controller that satisfies
Nominal performance with double requirement:

k(2Wp )Sok < 1
Robust stability with double uncertainty:

k(2Wu)T k < 1
Note: Other ratios are also possible. For example,

0 < < 1.
k|WpSo| + |(1 )WuTo|k < 1.
Choose to be large or small depending on which

of performance or robust stability can be more easily
satisfied.

Other Performance Specification?

Robust performance theorem is stated for performance
in terms of S. What about other performance
specifications? e.g.
Complementary sensitivity T (noise to output)

Input-disturbance sensitivity Si
Control sensitity Su (output disturbance to
control)?
Perturbed plant:
G(s) = (1 + G(s))Go(s); k()k < 1
Error sensitivity:
1
S(s) =
1 + To(s)G(s)
Perturbed sensitivities in terms of nominal:
S(s) = So(s)S(s)
T (s) = To(s)(1 + G(s))S(s)
Si(s) = Sio(s)(1 + G(s))S(s)
Su(s) = Suo(s)S(s)

A Conservative Result:
Design the nominal performance to be acceptable
Try to ensure that achieved performances are similar

to the nominal performance. This will be the case
if the error sensitivity
S(s) 1 + j0.
This is the case if |To(j)G(j)| <<< 1.
Hence, To(j) should roll off before G(j)

becomes significant, or
|To(j)Wu(j)| <<< 1
Notice that this condition is similar, but more

stringent than just robust stability. This result
means that to achieve robust performance when
performance is specified using sensitivities that are
not S, one can design the nominal system to have
good performance, and then make sure that the
complementary sensitivity has rolled off before the
uncertainty becomes important.

Servo-hydraulic Example
Consider a servo-valve controlled hydraulic actuator

with precisely controlled flow rate.
Nominal model based on incompressible fluid is just an

integrator:
X(s) 1
= Go(s) =
U (s) s
where x(t) represents the displacement, and u(t) is
the flow rate (normalized by the piston area) into the
actuator.
The fluid in the actual system has compressibility which

manifests itself as:
mx + bx + kf (x y) = 0;
y = u
This gives the transfer function from u to x to be:
K
G(s) =
s(ms2 + bs + K)

Thus, the multiplicative model error (MME) is:
G(s) Go(s)
G(s) =
Go(s)
(ms2 + bs) (s2 + 2wns)
= 2 = 2
m + bs + K s + 2wns + wn2
wn and depend on fluid compressibility which is highly

variable due to aeration, dirt, temperature, additives
etc.
Uncertainty: wn [200, 500], [0.1, 0.5].
Let us define bounds for uncertainty weights Wu(s)

so that
|G(j)| < |Wu(j)|
This should ensure that for any allowable G(s), we
can write it as:
G(s) = Wu(s)(s);
for some k(s)k < 1.

Performance Weight Wp
4
10
Wp = 100
(s + 20)5
2
10
0
10
9s
Wu =
2
10 s + 1200
4
10
6 9 s (s+200)
10 Wu =
(s + 1200)2
8
10
10
10
3 2 1 0 1 2 3 4
10 10 10 10 10 10 10 10
Various G(s) and 2 definitions of Wu

Since the bound of the uncertainty looks like a

20dB/decade rise flattening out at about 1500 Hz,
we first define
9s
Wu =
s + 1500
The value 9 is chosen so that it bounds the uncertainty
sufficiently. The bound is not very tight at low and
high frequencies. This may cause performance to be
conservative at low frequency.
Secondly try:
9s(s + 200)
Wu =
(s + 1500)2
This is motivated by the desire to lower the size in

low frequency portion, and to introduce a boost at
the resonance frequency. Thus we added a lead-lag in
the weighting. This turns out to be adequate for low
frequencies (at least for at least up to 1000 rad/s).

Next we define the performance criteria. The

specifications are:
Bandwidth of c = 20 rad/s
Output disturbance attenuation of 1/100 within

bandwidth.
Define performance weighting to be: wc = 20rad/s.
100wc5
Wp(s) =
(s + wc)5
The 100 is used, so that we have the 1/100

attenuation at D.C..
We used a high order filter to make sure that we dont

try too hard to achieve any performance above the
bandwidth wc.
The performance requirement is: for all ,
|Wp(j)S(j)| < 1
where S is sensitivity function.

Proportional control:
U (s) = K (R(s) X(s))
Nominal complementary sensitivity and sensitivity:
K s
To = , So = .
s+K s+K
Consider first nominal performance design.

Criteria is kSo(s)Wp(s)k < 1.
After several iterations, this is achieved by K = 600.
Nominal Performance WpS Robust stability WuT
2
10
WuT
0
WpS
10
2
10
4
10
6
10
8
10
10
10
1 0 1 2 3 4
10 10 10 10 10 10
Nominal design: K = 600
Nominal sensitivity and nominal complementary

sensitivity for K = 600

Unfortunately, K = 600 does not provide for robust

stability since kWuTok > 1.
Next, we try different Ks and apply the robust

performance criterion:
k|WuTo| + |WpSo|k < 1

K = 160
Sensitivity S Robust Stability WuT
0 0
10 10
2 2
10 10
4 4
10 10
6 6
10 10
5 0 5 5 0 5
10 10 10 10 10 10
Nominal Performance WpS Robust Performance WpS + WuT

10 2
10 10
0 0
10 10
10 2
10 10
20 4
10 10
5 0 5 5 0 5
10 10 10 10 10 10
Proportional Control K=160
K = 180
0 0
10 10
2 2
10 10
4 4
10 10
6 6
10 10
5 0 5 5 0 5
10 10 10 10 10 10

10 2
10 10
0 0
10 10
10 2
10 10
20 4
10 10
5 0 5 5 0 5
10 10 10 10 10 10

K = 200
0 2
10 10
0
10
2
10
2
10
4
10
4
10
6 6
10 10
5 0 5 5 0 5
10 10 10 10 10 10

10 2
10 10
0 0
10 10
10 2
10 10
20 4
10 10
5 0 5 5 0 5
10 10 10 10 10 10
K = 220
0 2
10 10
0
10
2
10
2
10
4
10
4
10
6 6
10 10
5 0 5 5 0 5
10 10 10 10 10 10

10 2
10 10
0 0
10 10
10 2
10 10
20 4
10 10
5 0 5 5 0 5
10 10 10 10 10 10

Gain - K kWuTok kWpSok kWuTo + WpSok

160.0000 0.8670 3.4806 3.5330
180.0000 0.9643 3.0948 3.1503
200.0000 1.0581 2.7859 2.8437
220.0000 1.1484 2.5330 2.5926
240.0000 1.2350 2.3222 2.3831
260.0000 1.3178 2.1437 2.2058
280.0000 1.4071 1.9907 2.0537
300.0000 1.4943 1.8581 1.9218
Conclusion: Proportional control cannot simultaneously
provide robust stability and performance.
Options:
Reduce performance specification - e.g. reduce

bandwidth wc
Reduce allowable size of uncertainties - e.g. better

system identification to nail down fluid / structure
natural frequency and damping ratio n, , thus
reducing Wu.
Use a different type controller loop shaping.

K = 180.
Robust Stability WuT
1
10
2 3 4
10 10 10
Nominal Performance WpS
0
10
2
10
4
10
0 1 2
10 10 10
K = 180
Sensitivity peaks at 15 rad/s whereas robust stability

peaks at 1500 rad/s.
Can we design a controller that only modifies the

sensitivities around 5 to 50 rad/s, but have little effect
at other frequencies?

Loop Shaping Principles
The robust performance control problem is essentially

a tradeoff between minimizing kWpSok (performance)
and kWuTok (robustness). As we saw, at any
frequency, either |Wp| or |Wu| must be less than 1.
Typically, at low frequency, performance requirement
is important and system uncertainty is low; and at high
frequency, the reverse is true. Thus,
At low frequency, |Wp| >> 1 and |Wu| << 1.
At high frequency |Wp| << 1 and |Wu| is large.
Thus intuitively, we solve |WpSo| < 1 at low frequency,

and |WuTo| < 1 at high frequency.
Let Lo(s) = Go(s)C(s). These requirements translate

to:
Low frequency (Wp dominates):

Wp
1 + Lo < 1 |Lo| > |Wp|


High frequency (Wu >> 1 dominates):

WuLo
< 1 |Lo| < 1
1 + Lo |Wu|
This provides some guidelines for designing Lo(s) to

satisfy robust performance.
To be more precise, we need to find bounds based on
the robust performance criterion itself.
We consider the cases of 1) when |Wp| < 1 (high
frequency) and 2) when |Wu| < 1 (low frequency) and
determine the necessary and sufficient conditions for
robust performance.
Case 1: |Wp| < 1: (High frequency - uncertainty is
important)
Sufficient condition:
1 |Wp|
|Lo| < |WpSo| + |WuTo| < 1.
1 + |Wu|
Necessary condition:
1 |Wp|
|Lo| < |WpSo| + |WuTo| < 1.
|Wu| 1

If |Wu| >> 1, both conditions approach,
1 |Wp|
|Lo| < (25)
|Wu|
Case 2: |Wu| < 1: (Low frequency - performance is

important)
Sufficient condition:
|Wp| + 1
|Lo| > |WpSo| + |WuTo| < 1.
1 |Wu|
Necessary condition:
|Wp| 1
|Lo| > |WpSo| + |WuTo| < 1.
1 |Wu|
If |Wp| >> 1, both conditions approach,
|Wp|
|Lo| > (26)
1 |Wu|
These bounds (25)-(26) determine the design rule for

Lo(s) = C(s)Go(s).

Loop Shaping Procedure
Setup:
Open loop plant G(s) is stable, and minimum phase

(no RHP zeros)
Wu(s) and Wp(s) are designed such that
min{|Wu(j)|, |Wp(j)|} < 1,
Procedure:
1. Plot on log-log sacle, magnitude versus frequency

(25)-(26):
At frequencies where |Wp| > 1 > |Wu| (low
frequencies):
|Wp|
,
1 |Wu|
At frequencies where |Wu| > 1 > |Wp| (high
frequencies):
1 |Wp|
|Wu|

2. Construct Lo(s) = Go(s)C(s) such that |Lo(j)|

within the required bounds
|Lo| > low frequency bound
|Lo| < high frequency bound
3. Choose Lo(s) such that at Lo(j) passes through

|Lo(j)| = 1 with gentle slope (-20db/decade or
-40dB/decade). This determines the phase margin.
4. Roll off Lo(s) (at least) as fast as Go(s) so that

C(s) is proper.
5. Check robust performance - |WpSo|+|WuTo| < 1.
6. Check nominal stability: roots of 1 + Lo(s) = 0

should lie on open LHP.
7. Determine C(s) from the Lo(s).

One possibility is to construct a nice looking Lo(s)

first, and then take
C(s) = Lo(s)/Go(s)
Another possibility is to start with L0o(s) = kG(s) and

then successively modify,
L1o(s) = kGo(s) L2o(s) = kC1(s)Go(s)

L3o(s) = kC2(s)C1(s)Go(s) . . .
where Ci(s) are typically lead-lag controller,
i (s + i)
Ci(s) = .
i s + i
The controller is then
C(s) = kCm(s)Cm1(s) . . . C1(s).
Cross-over region can be tricky to ensure robust

performance is achieved.

Loop shaping example: EH actuator
Recall that using proportional control, it is not possible

to have both adequate performance and robustness at
the same time. However, for C(s) = K = 180,
Performance curve |WpSo| peaks at 10 15 rad/s
Robust stability |WuTo| peaks at 1500 rad/s with

adequate margin.
Thus, it seems feasible that if we can improve |WpSo|

at around 10-15 rad/s without disturbing |WuTo| at
high frequencies too much, robust performane can be
achieved.
We use loop shaping techniques to guide us.

Trial 0: (nominal proportional controller)
180
L(s) = C(s)Go(s) =
s
EH actuator example Proportional Control (K=180)
4
10
3
10
2
10
1
10
0
10
1
10
2
10
1 0 1 2 3 4
10 10 10 10 10 10
As expected, this controller fails in the performance

bound.

Trial 1: Boost gain between 1 to 10 rad/s
s+1 s+1
Lo(s) = 180 Go(s) = 180
(s/10 + 1) s(s/10 + 1)
EH actuator example Proportional + lead

4
10
3
10
2
10
1
10
0
10
1
10
2
10
1 0 1 2 3 4
10 10 10 10 10 10
Performance bound is satisfied at the expense of

violating the robustness bound. This is verified
by plotting the robustness stability and performance
curves.


0 2
10 10
1 0
10 10
2 2
10 10
3 4
10 10
4 6
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10

0 1
10 10
5 0
10 10
10 1
10 10
15 2
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10
Robust stability, nominal performance, robust

performance curves for
s+1
C(s) = 180
s/10 + 1

Trial 2: Reduce gain at high frequency to regain

robustness
s/100 + 1 s+1
Lo(s) = 180 Go(s)
(s/10 + 1) s/10 + 1
s/100 + 1
= 180
s(s/10 + 1)2
4
10
3
10
Trial 1
2
10
Trial 0
1
10
Trial 2
0
10
1
10
2
10
1 0 1 2 3 4
10 10 10 10 10 10
Both performance and robustness bounds are satisfied.

Need to check robust performance curve.


0 0
10 10
1
10
2
10
3
10
4 5
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10

0 0
10 10
5
10
1
10
10
10
15 2
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10
Robust stability, nominal performance, robust

performance curves for
(s + 1)(s/100 + 1)
C(s) = 180
(s/10 + 1)2
Robust performance is satisfied.
Make sure to check that the system is nominally stable.

The characteristic polynomial is:
180(s + 1)(s/100 + 1) + s(s/10 + 1)2
Roots are: 99.5 90.5j and 0.996. Thus the

system is stable.
Nyquist plot confirms that the nominal loop is stable.

Thus, by robust stability, the system is robustly stable
as well.
Nyquist Diagram
10
5
Imaginary Axis
10
15
20
1.2 1 0.8 0.6 0.4 0.2 0

Real Axis

Control design via innovation feedback
Recall that using observer state feedback,
x = Ax + Bu L(C x y)
u = K x
the controller itself satisfies [See Goodwin pp. 512 for

proof]:
L(s) P (s)
U (s) = Y (s) + V (s) (27)
E(s) E(s)
where
L(s) det(sI A + LC + BK)

= 1 + KT1(s) =
E(s) E(s)
P (s) KAdj(sI A)J
= KT2(s) =
E(s) E(s)
P (s)
= K(sI A + LC + BK)1L
L(s)

Controller can be written as a two degree of freedom

controller form:

E(s) P (s)
U (s) = V (s) Y (s)
L(s) E(s)
Or as a 1 degree of freedom controller form:
P (s)
U (s) = (R(s) Y (s))
E(s)
P (s)
where V (s) = E(s) R(s).
The innovation is the output prediction error:
:= y C x = Ce
Therefore,
(s) = Y (s) C X(s)

= Y (s) CT1(s)U (s) CT2(s)Y (s)
= (1 CTs(s))Y (s) CT1(s)U (s)
where
T1(s) = (sI A + LC)1B

T2(s) = (sI A + LC)1L

In transfer function form:
L(s) P (s)
U (s) = Y (s)
E(s) E(s)
Bo(s) CAdj(sI A)B

G(s) = =
Ao(s) det(sI A)
E(s) = det(sI A + LC)
F (s) = det(sI A + BK)
L(s) = det(sI A + LC + BK)
P (s) = KAdj(sI A)L
P (s) 1
= K [sI A + LC + BK] L
L(s)
Then, it can be shown (see Goodwin P545) that the

innovation
Ao(s) Bo(s)
(s) = Y (s) U (s)
E(s) E(s)
Consider now that observer state feedback augmented

with innovation feedback,
u = v K x Qu(s)
where Qu(s) is filtered by the stable filter Qu(s)

(to be designed). Then,

L(s) P (s) B(s) A(s)
U (s) = Y (s)+Qu(s) U (s) Y (s)
E(s) E(s) E(s) E(s)
The controller transfer function becomes then:

P (s) A(s)
E(s) + Qu(s) E(s)
C(s) = L(s)
E(s) Qu(s) B(s)
E(s)
The nominal sensitivity functions, which define the

robustness and performance criteria, are modified
affinely by Qu(s):
Ao(s)L(s) Bo(s)Ao(s)
So(s) = Qu(s) (28)
E(s)F (s) E(s)F (s)
Bo(s)P (s) Bo(s)Ao(s)
To(s) = + Qu(s) (29)
E(s)F (s) E(s)F (s)

For plants that are open-loop stable with tolerable pole

locations, we can set K = 0 so that
F (s) = Ao(s)
L(s) = E(s)
P (s) = 0
so that
Bo(s)
So(s) = 1 Qu(s)
E(s)
Bo(s)
To(s) = Qu(s)
E(s)
In this case, it is common to use Q(s) := Qu(s) AE(s)

o (s)
to get the formulae:
So(s) = 1 Q(s)Go(s) (30)

To(s) = Q(s)Go(s) (31)
Thus the design of Qu(s) (or Q(s)) can be used to

directly influence the sensitivity functions.

For instance, using Eqs.(28)-(29):
Minimize sensitivity kWp(s)S(s)k for nominal

performance:
L(s)
Qu(s) = F1(s)
Bo(s)
Minimize complementary sensitivity kWu(s)T (s)k:
P(s)
Qu(s) = F2(s)
Ao(s)
where F1(s), F2(s) are close to 1 at frequencies where

kS(s)k and kT (s)k need to be decreased.
Similarly, using Eqs.(30)-(31) for stable open loop
systems:
Minimize sensitivity kWp(s)S(s)k for nominal

performance:
Q(s) = G1
o (s)F1 (s)
Minimize complementary sensitivity kWu(s)T (s)k:
Qu(s) = G1
o F2 (s)

where F1(s), F2(s) are close to 1 at frequencies where

kS(s)k and kT (s)k need to be decreased. Notice that
it is not possible to do both at the same time.
Remarks:
Generally, if F1(s) and F2(s) need to be active in

overlapping ranges, then the control design will not
be feasible.
The internal stability of the system is ensured if

Qu(s) is stable.
How Qu(s) should be designed need to be modified

in case when Bo(s) or Ao(s) are non-minimum
phase.

Electrohydraulic Actuator Example using

Qu(s) feedback
Simplified model for the EH system,
Bo(s) 1
G(s) = =
Ao(s) s
States space model:
x = u + di
y = x + do
where di and do are input and output disturbances.
The observer is:
x = u L(x y)
(s + L)x(s) = u(s) + Ly(s)

The innovation is:
= y y = y x
u + Ly
(s) = y(s)
s+L
s y(s) u(s)
=
s+L s+L
Let the observer state-feedback with innovation

feedback be:
u(s) = K x(s) Qu(s)(s)

[(s + K + L) Qu(s)]u(s) = (KL + Qu(s)s)y(s)
y(s) s(s + K + L) s
So(s) = = Qu(s)
do(s) (s + L)(s + K) (s + L)(s + K)
u(s) KL s
To(s) = = + Qu(s)
di(s) (s + K)(s + L) (s + L)(s + K)
This is consistent with the formulae (28)-(29) with

these definitions:
E(s) = s + L
F (s) = s + K
L(s) = (s + L + K)
P (s) = LK
First we consider K = 180. Without Qu(s) feedback,

this is robustly stable, but does not have the required
performance.
Nominal Performance
Bode Diagram 5
1 10
Magnitude (dB)
0
0.5 10
5
0 10
1
Phase (deg)
10
0.5 10
15
0
0 1
10
2 0 2 4
10 10 10 10 10 10
Frequency (rad/sec)
Robust stability Robust performance

0 1
10 10
0
10
1
10
2
10
5 3
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10
To improve nominal performance, we focus on

frequency below 200 rad/s:

>> wc1=200;
>> [B1,A1]=butter(2,wc1,s);
(s + L + K) B1(s)
Q(s) =
A1(s)
Nominal Performance
Bode Diagram 0
100 10
Magnitude (dB)
50
5
0 10
50
0
Phase (deg)
10
10
90
15
180
1 2 3 4 5
10
2 0 2 4
10 10 10 10 10 10 10 10 10
Frequency (rad/sec)

0 0
10 10
5 5
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10
This satisfies the robust performance criteria easily.
Lets try starting out with performance. Set K = 1200.

Without Qu(s), do not satisfy robust stability.

Nominal Performance
Bode Diagram 0
1 10
Magnitude (dB) 0.5 5

10
0
1
Phase (deg)
10
10
0.5
15
0
0 1
10
2 0 2 4
10 10 10 10 10 10
Frequency (rad/sec)

2
10
0
10
2
10
0
4 10
10
6
10
2 0 2 4 3
10 10 10 10 10
Clearly, this satisfies nominal performance but violates

robust stability in the region of [400, 4000] rad/s.
>> [B2,A2]=butter(2,[200,5000],s);
KLB2(s)
Q(s) =
sA2(s)

Nominal Performance
Bode Diagram 0
100 10
Magnitude (dB)
50
0 5
10
50
100
360
Phase (deg)
10
10
0
15
360
0 2 4 6
10
2 0 2 4
10 10 10 10 10 10 10 10
Frequency (rad/sec)

0 0
10 10
1
10
2
10
5 3
10 10
2 0 2 4 2 0 2 4
10 10 10 10 10 10 10 10
This also satisfies robust performance!

Constraints on Wp and Wu
Wp(s) and Wu(s) specify the desired performance and

allowable model uncertainties.
However, we cannot define Wp and Wu to be arbitrarily

large and expect that that a controller can be found
that solves the the robust performance problem. Here,
we show that they must respect each other, and respect
the limitations of the open-loop system.
Knowledge of these limitations help us define

meaningful performance specifications (Wp), and our
need to do accurate modeling (Wu).
First, some preliminary results ....

Preliminary Results
Maximum Modulus Theorem (MMT)
Let P (s) be a stable, rational transfer function. Then,
kP k := sup|P (j)| = sups with Re(s) 0 |P (s)|
It is obvious that LHS is less than or equal to the RHS.

The interesting part of this theorem is that they are
equal.
Consider L(s) = G(s)C(s). Let p be any pole of

L(s) and z be any zero of L(s), i.e. 1/L(p) = 0 and
L(z) = 0. We have:

1
S(p) = = 0; T (p) = 1 S(p) = 1;
1 + L(s) s=p

L(s)
T (z) = = 0; S(z) = 1 T (z) = 1;
1 + L(s) s=z

Constraint 1: Wu & Wp cannot simultaneously be

large
For the robust performance to be solvable, a necessary

condition is: for all ,
min{|Wu(j)| , |Wp(j)|} < 1.
Proof: Suppose that |Wu| |Wp| (reverse the

argument otherwise). At each ,
|Wu| = |Wu(S+T )| |WuS|+|WuT | |WuS|+|WpT |
Thus, |WuS|+|WpT | < 1 (robust performance) implies

that |Wu| < 1.
Significance: We cannot simultaneously tolerate

uncertainty, and expect good performance at any
frequencies.
One cannot have better than open-loop performance

(|S| < 1 as guranteed by |WpS| 1 and |Wp| > 1)
when uncertainty is larger than 100% (|Wu| > 1).

Constraint 2: Right Half Plane poles and zeros limit

robustness (Wu) and performance (Wp)
Suppose that C(s) internally stabilizes the nominal
plant Go(s). Let So and To be the nominal sensitivities,
and p and z are respectively an unstable pole, and a
non-minimum phase zero of Lo(s) = Go(s)C(s), i.e.
1/Lo(p) = 0 with Re(p) 0; and Lo(z) = 0 with
Re(z) 0.
The nominal (and robust) performance problem

cannot be solved if |Wp(z)| 1, since
kWpSok = supRe(s)0|WpSo|
|Wp(z)So(z)| = |Wp(z)|.
The robust stability (and robust performance)

problem cannot be solved if |Wu(p)| 1, since
kWuTok = supRe(s)0|WuTo|
|Wu(p)To(p)| = |Wu(p)|.
The remedy for this is to reduce performance

requirement (smaller |Wp(z)|), and better system
identification (lower uncertainty, and smaller |Wu(p)|).

The problem is more acute if there is a pair of RHP

pole and zero close to each other.
Consider the open loop system:
sz
G(s) = G1(s)
sp
where z and p are RHP zero and pole, G1(s) does not
have any RHP poles or zeros. Then, it can be shown
that:

z + p
kWpSok |Wp(z)|
z p

z + p
kWuTok |Wu(p)|
z p
Thus, as p z, the lower bounds for kWpSk and

kWuT k is significantly amplified. This is related to
the fact that the unstable pole is nearly canceled out by
the non-minimum phase zero. Thus, the unstable mode
becomes either nearly uncontrollable or unobservable.

Constraints due to Bode Integral

These constraints are sometimes referred to as the
principle of conservation of dirts or the area formula.
The general meaning is that for systems that satisfy the
conditions of the theorems, it is not possible to improve
the performance / robustness at all frequencies.
Theorem: (Bode Integral Theorem for Sensitivity)
Let the open loop system L(s) = G(s)C(s) have the
following properties:
It has relative degree (i.e. order of denominator

minus order of numerator) nr 1.
L(s) has M 0 RHP (unstable) poles (counting
multiplicity): p1, p2, . . . , pM , Re(pi) > 0.
Let := limssL(s). [Note: = 0 if nr 2]
1
Sensitivity function is given by S(s) = 1+L(s) .
Then, the sensitivity function S(j) satisfies:
Z M
X
ln|S(j)|d = Re(pi). nr > 1 (32)
0 i=1
Z M
X
ln|S(j)|d = + Re(pi). nr = 1.
0 2 i=1
(33)

Significance:
L(s) has relative degree nr > 1 if both plant G(s)

and controller C(s) are strictly proper. Then, (32)
applies.
Decreasing |S(j)| at some frequencies will

increase it at other frequencies. Hence, the dirt
is conserved.
The total amount of dirt is increased if the open-

loop system L(s) is unstable since the RHS of
(32)-(33) is increased.
When nr = 1, the total amount of dirt is

decreased by increasing the high frequency gain
(e.g. L(s) = K/s).
The sensitivity peak kSk which is inversely related

to robustness may be increased.

Example: Proportional control of 2nd order plant with

no zeros.
k
L(s) = kG(s) = 2
;
s +s+1
s2 + s + 1
S(s) = 2
s + s + (1 + k)
log S(jw)
1
10
L(s) = k 1
2
s +s+1
k=0
0
10
k=10
1
10
2
10
0 1 2 3 4 5 6 7 8 9 10
rad/s
Note that as k increases, the sensitivity at low

frequencies decreases but the peaks increase at other
frequencies.

Theorem: (Bode Integral Theorem for Complementary

Sensitivity) Let the open loop system L(s) = G(s)C(s)
have the following properties:
L(s) has at least 1 pole at 0 (i.e. L1(0) = 0 or

T (0) = 1).
L(s) has M 0 RHP (non-minimum phase) zeros
(counting multiplicity): c1, c2, . . . , cM , Re(ci) > 0.
kv is the velocity constant - i.e.
1
dT (s)
kv = lims0 = lims0sL(s)
ds
L(s)
Complementary sensitity is given by T (s) = 1+L(s) .
Then, the sensitivity function T (j) satisfies:
Z M
1 X 1
ln|T (j)|d = (34)
0 2 c
i=1 i
2kv

Significance:
If L(s) has at least 2 free integrators (poles at 0),

then kv = . This ensures that steady error is 0
for ramp input.
Decreasing |T (j)| at some frequencies will

increase it at other frequencies. Hence, the similar
to the S(s) story, dirt is conserved.
The total amount of dirt is increased if the open-

loop system L(s) has non-minimum phase zeros,
since the RHS of (34) is increased.
If L(s) has only 1 free integrator, one can decrease

the total dirt by tolerating steady state error due to
ramp input.
If L(s) does not have free integrators (i.e. L1(0) 6=

0 or T (0) 6= 1, then a similar relation to (34) exists,
except that in the integral we have ln |T (j)
|T (0)| . This
however, does not pose limitation on making T (j)
small in all frequencies.

Application: Closed loop bandwidth and

Open loop unstable pole
Constraints on Wp, Wu imply the following design rule:
The closed loop bandwidth should be larger than the

magnitude of the unstable open loop pole
Usually a factor of 2 is used.
Reasoning 1: (From complementary sensitivity

function)
Suppose that the uncertainty weighting is chosen so

that it is important at high frequency, unimportant at
low frequencies. This roughly translates to requirement
on To since kWuTok < 1.
s
One possibility is Wu(s) = o + T1 . So that
T o
Wu1 = ; |To(jo)| |Wu1(jo)|.
(sT + )
Hence, the cross over frequency is o, and Wu(0) =

T 1. Here, we can interpret o as the bandwidth of
the system, since beyond which, we allow To(j) to
be small.

Let p be a real unstable pole of the open loop

plant Go(s).
Then, from the constraint that |Wu(p)| < 1,
|p| 1 p
+ < 1 o > |p|.
o T 1 1/T
If T is 2 (50% uncertainty at D.C.), then this gives the

rule of thumb with the factor of 2.
Reasoning 2: (From sensitivity function)
Let Go(s) have a real unstable pole at p , and

Lo(s) = Go(s)C(s).
As a rough approximation, let the open loop gain

|Lo(j)| 0 when > o where o is the bandwidth
of the closed loop system. This implies that So(j)
1 (ln|So(j)| 0) when > o. Let M be the max
of |So(j)| (sensitivity peak)
From Bode integral (32),

Z o Z
p = ln|So(j)|d + ln|So(j)|d oln(M )
0 o

This shows that the sensitivity peak M ep/o .

Since M should be reasonably small (otherwise, the
system will behave much worse than open loop), p/o
should not be large. In particular, if o = p, then
M = e 23 which is unacceptable in most cases.
When, o = 2p, the estimated lower bound for M is
4.81.
Note: M >> 1 is bad also for robustness, since

T = 1 S. Thus, large T requires very good modeling
effort to maintain stability. Also, recall |So1(j)| is
distance of Lo(j) to (1, 0).

Application: Closed loop bandwidth and

Open loop non-minimum phase zero
The constraints on Wu and Wp also imply the following

design rule:
The closed loop bandwidth should be smaller than

the magnitude of the non-minimum phase zero
Suppose that the performance weighting Wp(s)

satisfies:
1 s
Wp = S
s + S
Thus, the sensitivity becomes important for < o;
high frequency sensitivity requirement is given by S.
Let z be a real non-minimum phase (RHP) zero of

Go(s). Then from the necessary condition, |Wp(z)| <
1,
z + oS
< 1 o < (1 1/S)z < z
Sz

Limitations - summary
Wu and Wp cannot be large simultaneously.
Open loop RHP poles and zeros limit how Wu and

Wp can be defined
Conservation of dirt theorems say that improving

sensitivity So(j) in some frequencies require
payment at others.
Similar theorem for To(j), especially when infinite

open loop D.C. gain (open loop integrators).
Design implications:
Closed loop pole should be faster than open loop
unstable pole;
Closed loop pole should be slower than open loop
non-minimum phase zero.
Problem when open loop pole is fast, and non-
minimum phase zero is slow consider changing
system architecture.

Example: Inverted pendulum
Balancing a beam on a palm. Let u be the force on

the palm and M is its mass; m and l are the mass and
the length of the beam. Assume that the beam mass
is concentrated at the tip. Let y be the tip position
and x be the position of the palm.
Transfer function from u to x:
ls2 g
Gux(s) = 2
s (M ls2 (M + m)g)
q
This has an unstable pole at + (MM
+m)g
l , and a non-
p
minimum phase zero at + g/l.
Transfer function from u to y:
g
Guy (s) =
s2(M ls2 (M + m)g)
This does not have any zeros.

Conclusions:
Control is more difficult as the beam becomes

shorter, or if the beam is heavier, as the unstable
pole gets larger. This requires faster bandwidth on
the part of the human controller.
Control by looking at the palm (x) is virtually

impossible, especially for short beams, because of
the unstable pole is faster than the non-minimum
phase zero. Thus, there will inevitably be a
large sensitivity peak (kSok) and hence robustness
problems.
Control by looking at the tip of the beam (y)

is easier since there are no zeros. For l = 0.5m,
assume that the palm is much heavier than the beam
(M >> m). Using a factor of 2 p the required
bandwidth is 1.4 Hz. This is quite reasonable for
humans.

Robust Stability and Performance: ME8281 - Last Updated: April 21, 2008 (C) Perry Li

Uploaded by

Copyright:

Available Formats

Robust Stability and Performance: ME8281 - Last Updated: April 21, 2008 (C) Perry Li

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Robust Stability and Performance: ME8281 - Last Updated: April 21, 2008 (C) Perry Li

Uploaded by

Copyright:

Available Formats

ME8281 - Last updated: April 21, 2008 (c) Perry Li

Robust stability and Performance

Topics: ([ author ] is supplementary source)

Sensitivities and internal stability (Goodwin 5.1-5.4)

Modeling Error and Model Uncertainty (Goodwin

Robust stability (Goodwin 5.7-5.10, Doyle 4.2)

Robust performance (Doyle 4.3-4.4, Goodwin 5.9)

Loop-shaping technique (Doyle 7.1) [We will cover

Innovation feedback and affine parameterization

The following topic(s) are for your information:

Performance limitation (Glad 7.3-7.4, Doyle 6.1-6.2,

M.E. University of Minnesota 137

Nominal Sensitivity Functions

To - Complementary sensitivity (goal: small for

So - Sensitivity (goal: small)

Sio - Input disturbance sensitivity (goal: small)

Suo - Control sensitivity (goal: small)

M.E. University of Minnesota 138

The nominal loop is internally stable if and only if all

Let C(s) = P (s)/L(s), and Go(s) = Bo(s)/Ao(s).

Proposition The system is internally stable if and only

Ao(s)L(s) + Bo(s)P (s) = 0

all lie in the open left half plane.

M.E. University of Minnesota 139

where pole-zero cancellations at s = 2 and s = 0

Complementary sensitivity is stable:

Sensitivity is also stable:

However, control sensitivity is marginally stable:

U (s) C(s) (s + 2)(s + 4)

And, input disturbance sensitivity is unstable:

M.E. University of Minnesota 140

The effect of unstable pole at s = 2 shows up in

The effect of unstable pole at s = 0 shows up in

Characteristic equation has unstable and marginally

(s + 2)s + (s + 4)(s + 2)s = (s + 2)(s + 5)s

predicting correctly that the system is NOT

M.E. University of Minnesota 141

For linear systems, if Go(s) is the nominal system, and

G(s) = G(s) Go(s)

Go(s)G(s) = G(s) Go(s)

M.E. University of Minnesota 142

Example: Time delays -

The transfer function of time delay is e s. Since it is

The additive modeling error is:

The multiplicative modeling error is:

M.E. University of Minnesota 143

Multiplicative Modeling error of approximation of various orders for 1s time delay

Uncertain pole location:

1/(s + Am) 1/(s + A) where A [A0, A1]

Neglected (possibly structural) dynamics:

Neglected compressibility effect, etc...

M.E. University of Minnesota 144

Let Go(s) be the nominal plant. Consider a family of

G(s) = Wu(s)(s) is the multiplicative model

Wu(s) is a given stable uncertainty weighting

(s) is the uncertainty itself, which is any stable

The question of robust stability is whether a controller

M.E. University of Minnesota 145

Robust Stability Theorem: Assume that the