Telecommun Syst (2013) 53:107–113
DOI 10.1007/s11235-013-9683-4
HMM based estimation of NCS network load and its application
in switching control
Yan Wang ·Longfei Tian
Published online: 20 June 2013
© Springer Science+Business Media New York 2013
Abstract The analysis and design of the Networked Con-
trol System (NCS) are more challenging due to the random
time delays brought by the network. And modeling the un-
certain time delays has been becoming a significant focus of
research. This paper modeled the network delays with Hid-
den Markov Model (HMM), and then on the basis of Viterbi
algorithm, proposed a method to estimate the network’s cur-
rent load state using random time delay sequence. In further
simulation we first designed PID controllers corresponding
to the network’s different load states respectively, and ap-
plied the proposed method to guide switching control of the
networked closed-loop system. The results showed the ef-
fectiveness of HMM in network state estimation and switch-
ing control.
Keywords Hidden Markov Model · Networked control
system · Viterbi algorithm · Time delay estimation ·
Switching control
1 Introduction
Networked control system (NCS) is a feedback control sys-
tem in which the closed-loop is connected by network. NCS
has many advantages in comparison to the conventional con-
trol system, such as resource sharing, remote control, sim-
plicity in installation and maintenance, etc. However, the
random time delay of the network poses great effects on
Y. Wa n g (
) · L. Tian
School of Automation Science and Electrical Engineering,
Beihang University, Beijing, 100191, China
e-mail: w-yan@buaa.edu.cn
L. Tian
e-mail: wei_zhi_2009@sina.com
the stability and dynamic performance of the system, which
makes the analysis and design of NCS more complicated.
The problem of modeling the network time delay and de-
signing controller based on the time delay model has been
an intense area of recent research.
There have been many results on network time delay
modeling. For example, Nilsson [1] represented the NCS
with random time delay by linear random system, and con-
verted the influence of network induced time delay into a
LQG problem. Recently the Hidden Markov Model (HMM)
has drawn much attention for it can describe the relation-
ship between the network time delay and the network load
state. Elliott [2] proposed the recognition problem of NCS
states and transferring probabilities using HMM. Azzouzi
and Nabney [3] predict the variable network time delay
using HMM. Salamation and Vaton [4] model NCS with
package drops as a Hidden Markov Model. Besides, some
other studies [5–9] estimate the random time delays and
compensate the random delays as the control system model
is given. It can be seen that many advances have been
made in the study of time delay modeling in NCS. But
further work is still needed to be done in the application
of HMM to the analysis and design of NCS [10]. And till
now there is not much HMM based study on the network
load state estimation and the corresponding control strat-
egy.
This paper first models the uncertain time delay of the
NCS with the Hidden Markov Model. And then on the basis
of the model, a method estimating the current network load
state is proposed. Further, the switching control of the sys-
tem is implemented based on the estimated load state. The
remainder of the paper is structured as follows. Section 2 in-
troduces the HMM and Viterbi algorithm briefly. Section 3
models the random time delay of NCS with the HMM and
proposes a method for the on-line estimation of current net-