The Fusion Methods of Multi-sensor System based on
Pseudo-measurement Model Library
Haihong Ye
1
and Chenglin Wen
1
and Xiaoliang Feng
2
Abstract— For single sensor system, information transmitted
in wireless networks will appear the phenomenon of delay, out-
of-sequence even dropout, so that the process center cannot re-
ceive and handle with the information promptly and effectively.
Similarly, this situation also exists in the multi-sensor system
where the information needs to be transported by wireless
networks, even more severe. As a result, the complexity of
the information processing methods for multi-sensor systems
is more severe. The core difficulty of these methods is how
to deal with the information with the phenomena mentioned
above, that not only need to consider the coupling of the sensor
itself, but also need to consider the underlying coupling between
different sensors. In order to deal with the delay measurements
in multi-sensor systems, this paper builds a novel pseudo-
measurement model library which is corresponding to multi-
sensor system, then proposes three fusion filtering algorithms
to deal with the networked measurements accurately in real
time: matrix fusion filter algorithm, centralized filter algorithm
and sequential filter algorithm. Meanwhile, the final simulation
examples demonstrate the effectiveness of the three proposed
methods.
Index Terms— Wireless sensor network; Delay measurement;
Pseudo-measurement model library; Matrix fusion filter algo-
rithm; Centralized filter algorithm; Sequential filter algorithm.
I. INTRODUCTION
In recent years, as the rapid development of network
technology, especially the wireless network technology and
low-power wireless sensor network technology, people con-
stitute a networked multi-sensor information fusion system,
which introduces wireless communication network as a hub
of information transmission[1-3].This networked system with
high scalability and maintainability, which can reduce system
wiring, enhance system flexibility and so on. However,
the information transmitted in the wireless sensor network
(WSN) can hardly avoid the delay, loss, and other uncertain
phenomena, due to the fact that the bandwidth of WSN
is limited, each node obtains the network resources by
competition. On the other hand, the wireless sensor network
is a self-organizing network constituted by a large number of
sensors in the monitoring area, aimed at cooperative sensing,
collecting and processing the information of detection object
within the network coverage area. Therefore, it is a key
link to perceive the physical world accurately in real time,
*This work was supported by National Nature Science Foundation under
Grant 61304258,61273075,61173133,61371064.
1
Haihong Ye, Chenglin Wen are with College of Automation,
Hangzhou Dianzi University, 310018 China Hangzhou, The Netherlands
yezi08061717@163.com; wencl@hdu.edu.cn;
2
Xiaoliang Feng are with College of Electrical Engineering, Henan
university of technolege, 450001 China Zhengzhou, The Netherlands
fengxl2002@163.com
through multi-sensor information fusion methods [4]. So the
fusion center client must design the fusion algorithms in
the case of limited communication power (eg: limited traffic
information, delay and packet loss) in WSN. Obviously,
communication limited will inevitably affect the estimation
performance, and the traditional multi-sensor data fusion
estimation theory cannot solve this problem. Therefore, there
exists an urgent need to develop fusion estimation theories
and algorithms under the circumstances of communication
constraints in WSN.
In recent years, according to the problems which exist in
wireless sensor networks, a large number of scholars have
been to conduct extensive research on those issues, such
as network packet loss rate, delay tolerant control method
[5], the delay measurement filter design [6] and its stability
analysis [7]. Where, for the issue of network latency, Y. Bar-
Shalom derived an A1 algorithm, which is corresponding
to one step optimal disorder measurement update[8]. W.
Zhou studied the optimality of A1 algorithm, which was
under the different process noise discrete methods[9], M.
Mallick put forward a B1 algorithm [10], X. Feng present
a delay filtering method which was based on the pseudo-
measurement delay measurement model library[11], Y. Bar-
Shalom proposed Al1 and Bl1 algorithm [12,13], all these
methods are studied based on a single sensor system, but
in practice, multi-sensor system is more widespread. So,
for a multi-sensor delay network, X. Shen gave the optimal
centralized update CA algorithm [14], the optimal distributed
update DA [15] algorithm and CAl, DAl algorithm, and
also gave CAl1, DAl1 algorithms, in the sense of linear
minimum variance. It is shown that the above multi-sensor
delay fusion filtering algorithms can not be operated until
all the delay measurements received by the fusion center.
This leads to weak real-time processing capability and high
storage requirements.
This article extends the idea of literature [11] to estab-
lish the pseudo-measurement model library for delay multi-
sensor system. Compared to the literature [11], the fusion
center overcomes more complexity to describe the pseudo-
measurement model library of the multi-sensor system. The
pseudo-measurement model library clearly shows the re-
lationship between the possible delay measurements and
current system state and the statistical property of relevant
parameters, which are usually consistent for stable systems.
Based on the pseudo-measurement model library, this paper
proposes three real-time fusion filtering algorithms: matrix
fusion algorithm, centralized fusion algorithm and sequential
fusion algorithm, in which, the matrix fusion algorithm only