Abstract—This short paper illustrates the use of empirical
gramians for controllability/observability analysis of nonlinear
systems and compares the extracted information to results
obtained from linear gramians and nonlinear observability
matrices. It is shown that empirical gramians can more
accurately represent controllability/observability of a
nonlinear system over an operating region than if linear
gramians are used. At the same time the information contained
in empirical gramians is easier to extract and interpret than if
Lie algebra-based controllability or observability matrices are
used.
I. INTRODUCTION
mpirical controllability and observability gramians have
been used for several years for model reduction of
nonlinear systems [1]-[3]. However, no results of using
empirical gramians for controllability and observability
analysis have been reported in the literature. This work
addresses this point by comparing observability information
from linear gramians [4], lie-algebra-based observability
matrices [5], and empirical gramians [1]. The presentation is
limited to observability due to the allocated space; however,
the results for controllability are similar in nature.
II. PRELIMINARIES
A. Linear Gramians
For a linear system of the form
Du Cx y
Bu Ax x
(1)
the linear observability (W
O,Linear
) and controllability
(W
C,Linear
) gramians [4]
dt Ce C e W
t A T t A
Linear O
T
0
,
dt e BB e W
t A T At
Linear C
T
0
, (2)
Manuscript received September 10, 2004.
A.K. Singh is a graduate student in the Department of Chemical
Engineering, Texas A&M University, College Station, TX 77843 USA
(email: abhay.singh@chemail.tamu.edu).
J. Hahn is an assistant professor in the Department of Chemical
Engineering, Texas A&M University, College Station, TX 77843 USA
(phone: 979-8453568; fax: 979-8456446; e-mail: hahn@tamu.edu).
can be used to compute observability and controllability of
states of the system.
B. Nonlinear Controllability and Observability Matrices
For nonlinear systems
) (
) ( ) (
x h y
u x g x f x
(3)
it is possible to determine if a system is locally observable or
controllable using concepts from differential geometry [5].
C. Empirical Gramians
Empirical gramians [1], [2]
p
i
r
l
s
m
ilm
m
C
dt t
rsc
W
1 1 1 0
2
) (
1
r
l
s
m
T
l
lm
l
m
O
dt T t T
rsc
W
1 1 0
2
) (
1
(4)
have been recently introduced for nonlinear systems of the
form of equation (3). Refer to [2] for the exact definition of
the variables from equation (4).
III. EMPIRICAL GRAMIANS FOR
CONTROLLABILITY/OBSERVABILITY ANALYSIS
Empirical gramians have been used for model reduction
of nonlinear systems and the benefit of using empirical
gramians over conventional linear gramians for model
reduction has been illustrated [2]. However, no current
results exist in the literature that directly compare the
information contained in empirical gramians to linear
gramians or nonlinear geometric methods for controllability
and observability analysis. An illustrative example is
presented in this section and a comparison of the
conclusions that can be drawn about observability from the
available information is made.
A. Illustrative Example
Consider the following nonlinear system
2
2
2 1 2 2
1 1
1 1 1
1
x y
x x x x
x x
(5)
which has an equilibrium point at (x
1
,x
2
) = (1,1). An
example without inputs is used due to the space constraint
and, therefore, only observability of the system will be
analyzed. The linearization of the system (5) at the
equilibrium point is given (in deviation variables) by:
On the Use of Empirical Gramians for
Controllability and Observability Analysis
Abhay K. Singh and Juergen Hahn
E
2005 American Control Conference
June 8-10, 2005. Portland, OR, USA
0-7803-9098-9/05/$25.00 ©2005 AACC
WeA05.1
140