![](bg2.png)
estimated as a sum of the relative position vectors.
Relative positioning is performed between all pair of
sensor nodes in the network and the optimal sum of
relative position vectors is selected to determine the
location of sensor nodes. To assess the performance
of our method, experiments have been conducted
using 53 wireless sensor nodes. The collected data
are analysed by using the proposed method in a post-
processing manner. The results show that the
success rate of estimating the relative positions is
considerably improved compared with the
conventional approach.
2 RELATED WORK
GPS-less localization method for large networks of
wireless sensor nodes has been intensively studied
by many researchers. For example, Bulusu et al.,
(2000) evaluates the effectiveness of a simple
connectivity metric method for localization in
outdoor environments. Moor et al., (2004) presents a
linear-time algorithm for localizing sensor network
nodes in the presence of range measurement noise
and demonstrates the algorithm on a physical
network. These results show that the accuracy
depends on the scale of distribution and is not high
enough for monitoring displacements of large civil
infrastructures, which needs a few centimetres to
sub-centimetres accuracy.
Many displacement monitoring systems using L1
GPS receivers are developed and demonstrated in a
real field. Gassner et al., (2002) developed a GPS-
based continuous monitoring system and applied it
to landslide monitoring. Shimizu, (2003) developed
a monitoring system and applied it to large open
quarries and landslide slopes. Seynat, et al., (2004)
developed the monitoring system, which uses a low
cost GPS receiver and a radio link, and applied their
system to volcano monitoring. These demonstrations
show that displacement monitoring using L1 GPS
receivers might be possible in terms of accuracy.
In order to deploy the sensor nodes densely
covering a large infrastructure, the cost for a single
observation point should be decreased further. So,
we have been developing the new displacement
monitoring system which combines the wireless
sensor network with an affordable L1 GPS receivers
connected to a small patch antenna which is
generally used for a mobile navigation (Saeki, 2008).
This combination enables us to decrease the cost of a
single observation point but brings about other
problems to be solved.
One of the problems is the energy consumption
of the GPS receiver. The sensor node of GWSN
keeps the GPS receiver off as long as possible to
save its battery. However, shortening observation
time results in the accuracy deterioration. To
overcome this problem, we have tried to develop a
new positioning method considering the condition of
dense sensor deployment.
The problem to determine the locations of many
GPS receiver simultaneously is known as network
adjustment (Han, 1995). In the network adjustment,
variance-covariance matrix between GPS sensors is
taken into account. However, it is so difficult to
estimate an appropriate variance-covariance matrix
in the application of a large infrastructure that
network adjustment might not be applicable.
3 GPS WIRELESS SENSOR
NETWORK
This section describes the outline of the present
system and the conventional relative positioning
method, and specifies the required technology.
3.1 Outline of the System
Figure 1 illustrates the schematic view of GWSN.
This system consists of a single central server and
many sensor nodes. The sensor node has a micro-
controller, a small wireless communication device, a
small battery and an affordable L1 GPS receiver.
The sensor nodes run according to the command
from the central server. After getting the command
to start observation, it turns on the GPS receiver
which outputs the raw binary messages to the micro-
controller every one second. The micro-controller
extracts the required data from the original binary
message and save them to the non-volatile memory.
The size of the original binary message is 266 bytes
and is compressed to 28 bytes in the present system.
After the sensor nodes collecting the data for several
minutes (e.g. 4 minutes), the central server orders
them to send their data back to itself. The locations
of the sensor nodes are analysed by the central
server in a post-processing manner.
In the prototype, a middle range series micro-
controller PIC16F877A (Microchip Technology
Inc.) is employed since any complex calculation is
not required. As an affordable L1 GPS receiver,
GT8032 (Furuno Electric co., ltd.) is used which is
capable of outputting L1 carrier phases in a Furuno
binary format. A small patch antenna is connected to
the receiver for saving cost. This kind of antenna is
PECCS 2012 - International Conference on Pervasive and Embedded Computing and Communication Systems
362