Continuous-Discrete Adaptive Observers For a Class of
Nonlinear Systems With Sampled Output
Guanglei Zhao
1
, Jie Mi
2
1. Institute of Electrical Engineering, Yanshan University, Qinhuangdao, 066004, China.
E-mail: glzhao517@126.com
2. Department of Human Resource, Yanshan University, Qinhuangdao, 066004, China.
E-mail: aimee@ysu.edu.cn
Abstract: This paper considers the continuous-discrete time adaptive observer design for a class of nonlinear systems with
unknown constant parameters and sampled output measurements. The proposed observer is in impulsive dynamics form with
an inter-sample output predictor, where the observer state flows continuously when the output is not available, and a corrective
term corresponding to measured samples is used to update the observer state. By assuming appropriate persistent excitation
conditions and following a technical lemma, an upper bound of the sampling intervals is derived, with which, the convergence of
the observer state and unknown parameters can be ensured. Finally, the proposed observer is used in an example of single-link
flexible-joint robot manipulator to show the effectiveness.
Key Words: nonlinear systems; continuous-discrete adaptive observer; sampled output measurements; persistent excitation;
unknown parameter
1 INTRODUCTION
State estimation is important in many engineering appli-
cations, such as tracking and navigation, secure communi-
cations, multi-agent systems, etc, this is due to the diffi-
culty or impossibility of measuring state variables in prac-
tice, therefore observer design problem for estimating non-
linear systems’ states has attracted lots of attention over the
last decades. Considerable effort has been devoted to de-
velop observer design methods for nonlinear systems, such
as [1],[2],[3],[4]. Moreover, for nonlinear systems with un-
known parameters, adaptive observers have been used to es-
timate both the states and unknown parameters in some lit-
erature [5],[6],[7].
One common feature of the above mentioned results is
the assumption that the system outputs are continuously
measured, however, in many real world applications, be-
cause digital sensors are used to measure signals, the out-
put measurements are generally available only at discrete
sampling instants. This leads to a continuous-discrete time
system, namely the system dynamics are continuous, but
for the observer, the output measurements are only avail-
able at discrete sampling instants. The observer design for
such continuous-discrete time systems can be traced back
to [8] who introduced a continuous-discrete Kalman filter to
solve a filtering problem for stochastic systems, inspired by
the results, high-gain observers was adapted to continuous-
discrete systems in [9]. Based on these early contribution-
s, various observer design techniques have been proposed
for continuous time nonlinear systems with sampled output
measurements (see [10],[11],[12],[13],[14]). For instance,
continuous-discrete time observer was designed in [11] for
state affine systems, where the observer states were updat-
ed in an impulsive form at sampling instants, and the results
This work is supported by National Natural Science Foundation of Chi-
na under Grant 61673335, 61603329, China Postdoctoral Science Foun-
dation 2016M601283, Natural Science Foundation of Hebei Province
F2017203145, Research Foundation of Yanshan University 15LGA007,
B964.
were extended for adaptive case in [12]. The cases of de-
layed measurements and time-delay systems with sampled
measurements were considered in [13] and [14] respective-
ly. Sampled-data adaptive observer for state affine output-
injection nonlinear systems was proposed in [15] and an ap-
proach of enlarging sampling intervals was presented in [16].
A promising approach to deal with the sampled measure-
ments case was proposed in [17], where a sampled-data non-
linear continuous time observer was designed coupled with
an inter-sample output predictor, and the robustness proper-
ties of the continuous time design can be inherited by the
sampled-data design with small sampling intervals. The out-
put predictor was used in [18] and [19] for triangular Lips-
chitz systems, and the observer states are continuous with-
out using corrective term at sampling instants. Another ap-
proach was described in [20] where an impulsive observ-
er was proposed for a class of uniformly observable sys-
tems with sampled outputs, the observer state was updated
with a correction term at discrete instants. Adaptive impul-
sive observers (AIOs) for nonlinear systems with unknown
parameters were proposed in [21], and the parameter esti-
mation law was improved in [22] by using an impulse-free
time-varying differential equation associated with the im-
pulse time sequence. However, continuous output measure-
ments are required in both [21] and [22] for parameter es-
timation. Recently, following the approach in [23] and [9],
the continuous-discrete observers in [24] are obtained in t-
wo steps: first, the state estimation is computed by integrat-
ing the model during sampling intervals, then, the observer
makes an impulsive correction to the estimation at sampling
instants. This two step approach was used in [25], where es-
timation error was a solution of an appropriate unknown lin-
ear parameter varying system, and a framer was constructed
such that the estimation error was ensured to be asymptoti-
cally stable. The proposed approaches in [24] and [25] were
effective tools for state estimation, but unknown parameter
estimation problem was not considered.
In this work, we consider continuous-discrete time ob-
server design for a class of nonlinear systems where the out-
Proceedings of the 36th Chinese Control Conference
Jul
26-28, 2017, Dalian, China
787