J Control Theory Appl 2009 7 (4) 427–432
DOI 10.1007/s11768-009-8071-9
Stability margin of uncertain control systems
Bin L
¨
U, Qinghe WU
(
Department of Automatic Control, Beijing Institute of Technology, Beijing 100081, China)
Abstract: Achieving stability is the essential issue in the control system design. In this paper, four approaches that can
be used to calculate the stability margin of the interval plant family are summarized and compared. The μ approach gives the
bounds of the stability margin, and good estimation can be obtained with the numerical method. The eigenvalue approach
yields accurate value, and the MATLAB’s function robuststab sometimes provides wrong results. Since the eigenvalue
approach is both accurate and computationally efficient, it is recommended for the calculation of the stability margin, while
utilization of the function robuststab should be avoided due to the unreliable results it gives.
Keywords: Interval plant family; Stability margin; Stability radius; MATLAB’s function robuststab
1 Introduction
A critical issue in the process of control system design
is to achieve stability. If the plant is free of unstable pole-
zero cancellation, there must exist a controller such that the
closed loop system is stable. However, this controller may
be unreliable, because the models of the system may be in-
accurate. In most situations, we consider the control system
shown in Fig.1, where P(s, δ
N
, δ
D
) is the plant to be con-
trolled, δ
N
, δ
D
are the uncertainty parameter vectors of the
plant, and C(s) is the controller. We can assume that the
plant to be controlled is a family of interval plants:
P(s, δ
N
, δ
D
)=
P (s):P (s)=
n
j=0
[b
−
j
,b
+
j
]s
j
n
i=0
[a
−
i
,a
+
i
]s
i
. (1)
Fig. 1 A control system.
The robust stabilization problem associated with
P(s, δ
N
, δ
D
) is to find a single proper controller C(s),
meaning both the structure and the parameters of C(s)
are invariant, such that the negative feedback system com-
posed of P(s, δ
N
, δ
D
), and C(s) is internally stable for all
P (s) ∈P(s, δ
N
, δ
D
). This problem has been studied for a
long time, and some celebrated results have been obtained.
Chapellat and Bhattacharyya showed that a given controller
stabilizes the whole interval plant family if it stabilizes its
32 edge plants [1]. Barmish et al. showed further that an
interval plant family is robustly stabilized by a first-order
compensator if and only if the compensator stabilizes six-
teen of the extreme plants [2].
The above approaches can only be used to check the
stabilization condition of controllers. A further need is to
calculate the stability margin of the uncertain system, so
that the quality of the controller can be evaluated. Several
approaches have been proposed for this purpose, such as
the structured singular value (μ) approach, the numerical
method, the analytical eigenvalue approach, and the MAT-
LAB 2007b’s function, robuststab. With the increasing ap-
plication of computers and the development of numerical
software, MATLAB is extensively used by researchers. The
results obtained with MATLAB are almost always accurate
and sound, which results in heavy dependence on MATLAB
for engineering applications. However, the function robust-
stab of MATLAB 2007b often gives erroneous, and even
sometimes unreasonable results. In this paper, we summa-
rize these approaches and present examples illustrating in-
correct results obtained with MATLAB.
2 Definition of stability radius
To this end, we represent the interval plant family
P(s, δ
N
, δ
D
) in an equivalent form. Define
b
j
=
b
+
j
+ b
−
j
2
,a
i
=
a
+
i
+ a
−
i
2
,
w
N,j
=
b
+
j
− b
−
j
2
,w
D,i
=
a
+
i
− a
−
i
2
,
then the coefficient intervals can be represented as
[b
−
j
,b
+
j
]=b
j
+ w
N,j
δ
N,j
, − 1 δ
N,j
1,
[a
−
i
,a
+
i
]=a
i
+ w
D,i
δ
D,i
, − 1 δ
D,i
1.
Hence,
P(s, δ
N
, δ
D
)=
N
0
(s)+Δ
N
(s)
D
0
(s)+Δ
D
(s)
, (2)
where
N
0
(s)=b
n
s
n
+ b
n−1
s
n−1
+ ···+ b
1
s + b
0
,
D
0
(s)=a
n
s
n
+ a
n−1
s
n−1
+ ···+ a
1
s + a
0
,
Received 24 April 2008; revised 19 November 2008.
This work was supported by the National Natural Science Foundation of China (No.69574003, 69904003) and the Research Fund for the Doctoral
Program of the Higher Education (RFDP) (No.1999000701), and was partly supported by the Advanced Weapons Research Supporting Fund
(No.YJ0267016).