W-TREMORS, a Wireless Monitoring System for

Earthquake Engineering
Emanuele Goldoni and Paolo Gamba
Department of Electronics
University of Pavia, Italy
Email: {name.surname}@unipv.it

AbstractWireless Sensor Network is an emerging technology

which can signicantly reduce the cost and the time needed to
monitor the conditions of civil structures. This work presents WTREMORS, a new wireless sensing system for high-frequency
distributed data acquisition. Our sensor network is based on
the low-power and low-data rate standard IEEE 802.15.4,
and we used inexpensive hardware. Nevertheless, the system
meets requirements imposed by Structural Health Monitoring
applications thanks to an efcient use of hardware resources
and adopting a novel communication protocol which optimize
bandwidth utilization. In addition, we developed a complete
software architecture to integrate the Wireless Sensor Network
with an existing measurement system. Our prototype WSN was
tested through shaking tests in a controlled environment to
validate the approach, and to identify problems which should
be addressed before using the system on real-world buildings.
Preliminary results show that our solution can effectively monitor
seismic events providing high reliability and good performances.

I. I NTRODUCTION
Wireless Sensor Networks (WSNs) are a new emerging
technology which enables dense and distributed monitoring
of structures and buildings. Compared to other existing solutions, a WSN can signicantly reduce the cost and the time
needed to setup a data acquisition system without the need
for wires and complex hardware. However, Structural Health
Monitoring (SHM) applications have unique features, such as
high-frequency sensing, strict clock synchronization and high
data-rates. Moreover, critical measurements are often needed
in near real-time if data is received after a certain delay, the
value is not useful anymore.
Related work on using WSNs for structural monitor includes
notable experiments like Wisden [1], Tenet [2] and the Golden
Gate Bridge deployment [3]. Despite a number of notable efforts, we believe there room for improvement in the design of a
cheap and real-time system able to acquire time-synchronized
and high-frequency data.
This work presents W-TREMORS (Wireless Tremors, vibRations and Earthquakes MOnitoRing System), a new wireless sensing system for high-frequency data acquisition based
on simple and low-cost hardware. In order to meet requirements that SHM imposes, our system is based on a newly developed protocol which optimize bandwidth utilization. Moreover, a complete and fully-functioning software architecture
is used to manage the sensing elements and to integrate the
WSN with a database or an existing expert system. Compared

978-1-4244-6276-6/10/$26.00 2010 IEEE

to other proposed systems, our solution has achieved higher

sampling rates and reliable data delivery using inexpensive
hardware, resulting in less than 80 Euro per node.
The remainder of this work is organized as follows: Section
2 briey describes the related work on the design of structural
monitoring systems, while Sections 3 and 4 describe respectively the details of the hardware and the software elements
of our monitoring system. In Section 5 we report experiences
from deploying and testing our system on the eld. Finally
in Section 6 we propose directions for future research and
improvements.
II. R ELATED W ORK
Wisden is a multi-hop wireless system for structuralresponse data acquisition designed at the University of Southern California [1]. Wisden incorporates a reliable multi-hop
data transport mechanism and a data time-stamping system
which does not require global synchronization. A rst implementation of the wireless network acquired data at 50 Hz, and
used a simple run-length encoding technique to detect and
compress period of inactivity. The deployment experiences
on a seismic test structure drove to a re-design of parts
of the system [2] - in order to preserve the delity of the
structures frequency response, the sampling frequency has
been increased to 200 Hz and an onset detection scheme
was introduced during the compression phase. The delivery
protocol of Wisden proved to be reliable and robust to data
losses, achieving 100% successfully delivered packets during
tests. However, latency is a fundamental scaling limitation for
this system: a continuous 1-minute shake required about 6
minutes to completely transmit them.
Similarly, Mechitov et al. [4] developed a wireless sensing
system for structural monitoring based on Mica2 motes. Their
implementation rst stores sensor data in the ash memory values are delivered to the processing station when the
acquisition stops. Thus, memory limitations on the sensor
node allowed for recording no more than 90 seconds of
single-axes acceleration samples acquired ad 250 KHz. Clock
synchronization was also achieved adapting the FSTP time
synchronization service to the sensor network, which ensured
proper synchronization for the target application.
The Golden Gate Bridge (GGB) deployment [3], [5] is one
of the largest experimental WSNs for structural monitoring.
An integrated hardware and software system for a scalable

SHM wireless network has been designed and implemented

using a dedicated accelerometer board, Mica-Z sensor nodes
and the TinyOS operating system. Additional software components has been implemented, including a protocol for reliable
command dissemination and data collection, and software
functions for data pipelining, jitter control and high-frequency
sampling. The experimental GGB deployment 64 nodes were
distributed over the main span of the structure, collecting
vibrations synchronously over a 46-hop network. The nominal
sampling was performed at 1 KHz, but the digitalized signal
was downsampled to 50 Hz by averaging. The network layer
of this system is able to reliably deliver measurements to the
central station, although data gathering from the network is
performed off-line: about 250,000 samples are buffered by
each node, and when the storage memory is lled up data
are collected from every node one by one. A complete cycle
of sampling and data collection for the full network would
take about 9-12 hours [3].
Finally, two additional recent experiments have been presented in [6], [7]. In the rst work [6], researchers used lowcost hardware to developed a WSN for real-time SHM. The
resulting sensor network supports real-time acquisition from
up to 20 wireless units with negligible data losses; however
the bandwidth of the nal estimation is limited, that is below
128 Hz. In the second work a fully-working custom system
has been deployed in Torre Aquila, a medieval tower in Trento
(Italy). Basing on 4 months of operation, the authors showed
the effectiveness of the designed WSN for assessing the towers
stability, as it proved to deliver data with loss ratios < 0.01%
and an estimated lifetime beyond one year [7].
Although a number of works have addressed specic issues
related with structural monitoring, to our best knowledge WTREMORS is the rst prototype WSN which can provide 1
KHz sampling rate and real-time data delivery using nodes
which cost less than 80 Euro each.
III. H ARDWARE D ESIGN
The W-TREMORS monitoring system consists of a number
of sensor nodes and a central device, which is responsible for
network coordination and data collection. Each sensor node
acquires vibrations from the structure through a MEMS accelerometer, and then transmits wirelessly the acquired values.
A. Sensor Interface and Signal Conditioning
The accelerometer used to acquire vibrations in WTREMORS is a Kistler 8330A3 [8], a single axis MEMSbased capacitive sensing element for measuring static acceleration up to 3g or low-level, low-frequency vibrations
at 2000Hz. A custom integrated circuit measures the mass
position and provides the feedback force to restore the mass
to the center position - the raw, unltered output is provided
as a voltage signal.
The sensor unit needs a bipolar DC supply between 6V
and 15V . We designed a dedicated circuit (Figure 1) to
generate the suitable voltages from two rechargeable 9V
nickel-metal hydride cells. The output of the batteries has

Fig. 1.

The accelerometer conditioning circuit.

been used both to power directly the sensor and to generate

the negative voltage using TIs TLE2071 JFET operational
ampliers [9].
In addition, the conditioning circuit shifts the output voltage
to match the input range of the analog-to-digital converter of
the microcontroller the ADC single-ended voltage inputs
refer to 0V .
B. Wireless Sensor Network Platform
Our monitoring system has been developed on the top of
the Arduino Diecimila board [10]. This board is equipped with
one Atmel AVR ATMega 168 microcontroller, which features
a 8-bit processor running at 16 MHz. The microcontroller
has 1024 bytes of SRAM, an EEPROM memory of 512
bytes and 15 KBytes of ash, plus 1 KB used to store the
bootloader. The ATMega 168 also contains an onboard 6channel analog-to-digital converter each channel is a 10bit successive approximation ADC with a resolution of 4.89
mV/bit, and it contains a sample-and-hold circuit to keep the
input voltage at a constant level during conversion. The overall
computational power provided by this architecture is sufcient
to equip a WSN node [11], and the processor is very lowpower demanding and easily programmable using C language.
The wireless interconnection between the nodes of the
network relies on Digi XBee radio modules, which embed
the Freescale MC13193 chip and an external omnidirectional
antenna. This transceiver is fully IEEE 802.15.4-compliant
and operates in the 2.4 GHz band [12]. These devices do
not require a networking layer to implement communication
the chip provides a transparent serial interface for the
underlying 802.15.4 MAC transmitting at a nominal data-rate
up to 115.200 Kbps.
The radio module and the main board are also enclosed in
plastic box and attached to the accelerometer interface and
the two 9V batteries. The resulting prototype sensor node is
shown in gure 2.
C. Network coordinator
The coordinator of the network collects data from the
various nodes, synchronizes them and stores the acquired data.
This central node is composed of two parts: a radio interface
and a mini-pc. The rst part is actually an USB-to-serial
converter unit with a XBee-compatible pinout: the function








  
Fig. 3.

Fig. 2.

A prototype W-TREMORS sensor node.

of this module is to give direct access to the radio interface

through a standard serial connection.
On the other hand, the mini computer is a Koala Nano PC
[13] equipped with a Vortex 300MHz processor. This device
has a maximum power consumption of 6 W the little amount
of energy needed to work makes the Nano PC a perfect choice
even in battery or solar-cell powered applications.
IV. S OFTWARE D ESIGN
Two different fully-featured software stacks have been developed: one for the WSN side and the other designed to run
on the network coordinator.
A. Sensor Network
1) Transmission protocol: The transmission of acquired
data in the W-TREMORS system is based on the PRISM
(Protocol for Real-tIme Synchronous Monitoring) protocol
[14], [15]. PRISM is a solution for real-time synchronous
acquisitions in wireless sensor networks developed by our research group. The protocol can be implemented both in singlehop topologies and in multi-hop networks, where routing paths
are logically organized on a perfect binary tree.
PRISM drops CSMA/CA in favor of a TDMA-like scheduling of transmissions [16] - this allows to avoid collisions
among transmissions assigning non-overlapping time slots. To
be more precise, our protocol adaptively creates a scheduling
with an optimal number of time slots, thus achieving higher
throughputs compared to TDMA protocols with xed framesize. Given the maximum sampling frequency of the sensor
network, PRISM calculates off-line the optimal scheduling of
transmissions. Then, before starting the acquisition process,
the resulting TDMA arrangement is broadcasted into the
network and the acquisition can start.
The underlying idea is that all nodes specically developed
for wireless sensor networks are provisioned with an analogto-digital converter circuitry, a processing unit and the communication module, all embedded in a simple chip without DMA.
The main drawback of this integration is the lack of support
for the concurrent acquisition of samples and the transmission
of data. To cope with this issue, PRISM exploits the idle time
between two consecutive sensors acquisitions for transmitting
data. This interval is quite long, compared to the transmission
time of a single value hence, it is possible to reduce the



 

 

Packet format of the PRISM protocol.

channel usage putting multiple samples in a single packet,

that is transmitting them into a burst. The packet structure
of a burst is quite simple: the rst data in the packet is the
absolute timestamp associated to the rst sample of the group;
then, the burst of n of consecutive samples follows and an
integrity check code concludes the packet. A similar approach
has been proposed also in [1].
In [15] we have shown that it is possible to dene the
maximum number of nodes, the requirements on the network
synchronization and the data-rate in a exible manner, in
order to accommodate for different types of acquisitions. The
analysis of the protocol resulted in an analytical relationship
between the network size and the sampling frequency of nodes.
Numerical results show that the IEEE 802.15.4 standard,
originally designed for low data-rate applications, could be
used in combination with PRISM to build star and multi-hop
networks of tens of nodes, all acquiring more than one hundred
of samples per second [15].
2) Synchronization protocol: Synchronization between sensors plays a crucial role for most of real-world monitoring
applications. However, providing aligned clocks is a complex
task due to various challenging internal and external factors,
such as drifts, non-linearity of timers or signal propagation
times [17].
In a star topology, W-TREMORS adopts a centralized inband synchronization mechanism in order to compensate clock
discrepancy among sensor nodes without requiring dedicated
hardware. In the network, either a reference node or the
gateway itself periodically broadcasts a clock signal to all
the sensor nodes during a dedicated time slot. This solution
features low execution time, low trafc overhead, and it also
provides accurate synchronization for most practical uses. The
basic conguration requires the reference node to send the a
message every hyper-period. However, if the synchronization
protocol takes longer to calibrate all clocks, x consecutive
time slots could be used for this tasks, leaving the remaining
N x intervals free for sensor nodes. Similarly, various nonconsecutive slots in the same hyper-period could be used
instead to repeat the process, hence increasing the synchronization frequency without affecting sensors sampling period.
Nevertheless, the approach is exible and it could be easily
extended using other well-established algorithms [17], [18]
or [19] during the dedicated time slot in place of a single
reference signal.
On the other hand, nodes synchronization in multi-hop
topologies could be achieved using different approaches. For
example, the reference node can broadcast periodically a clock

signal to all the sensor nodes during a dedicated time slots.

However, all the devices in the network must be able to
receive the timing messages the applicability of this solution
depends on both the maximum output power of the radio
and the physical distance between the reference node and the
farthest one. Otherwise, distributed the algorithms could be
used to establish a global timescale throughout the network
for example, the Timing-sync Protocol for Sensor Networks
(TPSN) [17] performs pair wise synchronization along the
edges of a hierarchically organized network through a proper
exchange of messages.
B. Data Collection and Storage
A dedicate layer running on the gateway provides an unique
interface for heterogeneous acquisition systems: it is responsible for acquiring data from one or more wired or wireless
sensing system. This layer is also responsible for presenting
the sensed values to the end user in a consistent and standard
way. We decided to split the functionality of this layer into
two parts: the process receiving data from the sensor network,
hereafter called the producer, and the element responsible
for data management and presentation, hereafter called the
consumer. Each producer of this layer has a technologydependent interface specically designed for the acquisition
system used. This implies that the output of a this layer
depends too on the specic behaviour, design choices and
constraints of the acquisition system. We developed a specic
and working code for the W-TREMORS system, although
different systems could be connected to the same gateway.
We choose to implement the output of this component as
les because of the need to integrate different acquisition systems without introducing additional computational overhead
on the Sensor layer. The producer needs only to save the raw
output in a le with a known sufx and an incremental prex
the changing prex avoids collisions between successive
outputs in the queue while the the sufx extension depends on
the specic producer, and allows to identify and parse correctly
the data inside the le. In case the sufx is not enough to parse
the le, for example because the data formatting is not standard
or optional extensions have been used, it is also possible
to provide additional information in the rst line of the le
after the special character %. Since this feature is producerand consumer- dependent, its format and meaning should be
dened time after time by the specic sensing applications.
The consumer receives data from the queue and then output
measurements printing raw values on the screen, or on a different device. It is also possible to adopt an appropriate interface
to format, convert or elaborate values before printing. Finally,
the producer can also connect to a DataBase Management
System (DBMS) and store processed values in a standard and
easily-accessible way which guarantees atomicity, consistency,
isolation, durability and concurrency. The extraction of data
is also simplied by the existence of the Structured Query
Language standard for relational databases and bindings with
programming languages.

The acquisition of measurements and the reading of the

acquired values can be seen as a typical programming problem,
which can be solved with the adoption of the classical spooler
approach for multi-processes data exchange. The exchange of
data les through the spooler is performed by a dedicated
software component, which hides the complexity of the underlying system, managing the output and input queues and
moving les among them in a transparent way.
If the produced and the consumer processes run on the same
physical machine, the duty of the communication is smoothless
les movement across queues is performed using atomic
system calls for renaming le in the lesystem tree. Otherwise,
a network connection is needed to interconnect the coordinator
with a remote queues. File exchange over a network connection can be performed using either an FTP or a raw Internet
Socket communication.
V. E XPERIMENTAL E VALUATION
The prototype wireless sensor network was tested on a simple steel structure provided by the staff of the European Centre
for Earthquake Engineering (EUCENTRE) [20]. Data collected from the accelerometers were stored on a local DBMS,
and used to investigate sampling accuracy, synchronizationrelated issues and reliability of the wireless network. Preliminary experiments show that our wireless system can effectively
monitor seismic events, meeting requirements set by the civil
engineers of EUCENTRE.
A. Accuracy
Multiple nodes have been used to acquire vibrations in realtime, sampling vibrations at 1 KHz during a shaking test.
In order to evaluate the accuracy of measurements, values
collected from a single node have been imported in Matlab
and the spectrum has been analyzed the resulting frequency
of 3.7 Hz provided by our wireless system matched the actual
value provided by the existing EUCENTRE wired acquisition
system. We replicated the test using different structures with
known vibrations ranging from about 1 to 5 Hz: the spectra
analysis always provided results with a maximum error of 0.07
Hz between the wireless and the wired system.
Then, we repeated the process for an additional accelerometer put on the same side of the structure, and we investigated
the synchronization of these two sensors. Firstly, we performed
an off-line experiment in order to measure to relative drift
between the hardware clocks of the nodes. Using an oscilloscope, we found that the relative worst-case drift between
two devices is approximately 1.7 ppm this means we have
a 1 ms discrepancy after 10 minutes of operation, which is
accurate enough for practical uses and can be corrected by
the synchronization protocol.
On the other hand, we found that the basic software timer
function originally used by our nodes was deeply affected by
instability errors. We rewrote the code in order to use more
accurate low-level timing functions, thus achieving sufcient
clock stability and accuracy. As shown in Figure 4, vibrations

C. Power consumption

2.5
1st Sensor
2nd Sensor

2
1.5

Acceleration [g]

1
0.5
0
0.5
1
1.5
2
2.5

Fig. 4.

0.5

1.5

2
Time [s]

2.5

3.5

Synchronization between two W-TREMORS sensors.

TABLE I
T OTAL PACKET LOSS PERCENTAGE WITH VARYING NETWORK SIZE .
Sampling frequency
1000 Hz
500 Hz
200 Hz
100 Hz

2 node
5.0
2.8
1.9
0.2

4 nodes
6.9
3.1
2.4
0.9

6 nodes
8.1
3.6
2.9
1.3

The consumption of stock Squibee node is extremely high

the non-optimal design of the Arduino board requires 30mA
to operate. However, a few changes in the design of the sensor
node brought to acceptable results disconnecting all the
leds, the unused input pins and the USB-to-Serial chip we
lowered the current consumption of the whole sensor node to
17 mA while acquiring and 5 mA in sleep state. Estimating
the lifetime of a node is a complex task, mainly due to
the non-linear behaviour of commercially available batteries
[1]. However, we found that the lifetime of a W-TREMORS
sensor node continuously acquiring samples is about 2 days
and half when powered by a 1200 mAh battery. In addition,
it is worth noting that accelerations sampling draws more
current. Hence, the consumption of a sensor node depends
also on sampling frequency reducing sampling frequency,
putting the node in sleep state when inactive and using a
more power-efcient MEMS accelerometer would extend the
system lifetime. In addition, we estimated that the design of a
trimmed-down, customized version of the sensing board could
eventually reduce the overall power consumption to about 6
mA, resulting in node dying after more than one week of
operation.
VI. C ONCLUSION

data acquired from the two sensors exhibit thigh synchronization. During our experiments, the average offset between the
two sensors was about 0.8 ms, with a maximum discrepancy
of 2 ms obtained only in one test.
All the experiments described above have been repeated
lowering the sampling frequencies to 500 Hz, 200 Hz and 100
Hz, which correspond to signicant frequencies according to
EUCENTREs researchers, obtaining similar results.
B. Reliability
We tested the system inside a building we considered two
adjacent 5 m 3 m rooms, each having 20 cm brick walls and
both opened into a common corridor. We placed one sensor
node per room at 1.5 m height, while the gateway has been
installed in the passageway. We also increased the number
of nodes per room to 2 and 3 all values were obtained
in real-time and, as shown in Table I, the packet-loss ratio
never fall below 91.9% even with 6 wireless nodes acquiring
synchronously and delivering data at the same time.
Experiments also conrmed the results obtained in [21]
the positioning of the antenna is crucial for the reliability of
the wireless communication. If the node is put next to the
oor or to a wall, that is 10 cm or less, we experienced a
packet loss-rate higher than 90%. On the other hand, we did
not found notable differences in the reliability of the system
putting nodes at 0.5 m, 1 m or 2 m height. We suspect that
the intensity of multi-path phenomena is strong enough to
corrupt the signal if the node is too close to the wall or the
oor, although more tests would be needed before drawing
any general conclusion on this issue.

Sampling rates required for structural sensing are challenging, and standard WSN technologies available nowadays
have signicant limits. Nevertheless, this work shows how
general-purpose and low-cost hardware for WSNs can be used
to effectively monitor seismic events in real-time with high
accuracy and good reliability.
We consider this work only as the beginning of a larger
project: we intend to improve the design of our sensing
elements, implementing a fully working full-scale node and
then extending the experimental evaluation to a more complex
scenario. Moreover, power consumption on nodes will be
addressed deeply from both a theoretical and practical point
of view. Finally, we plan to introduce hybrid-ARQ schemes
and different channel-assignment mechanisms to improve the
reliability of transmissions.
Additional implementation details are all available at http://netlab-mn.unipv.it/wsn/ - future developments and reports will
be published in the same web page.
ACKNOWLEDGMENTS
We are grateful to Marco Beltrame, who helped us during
various stages of the implementation and testing of the testbed.
Special thanks to the staff and management of the European
Centre for Earthquake Engineering (EUCENTRE), in particular to Prof. Alberto Pavese an Dr. Filippo Dacarro for their
close cooperation with us and for providing access to their
shaking table and the wired acquisition system currently in
use. This work was partially supported by the STEP - Strategies and Tools for Early Post earthquake assessment project,
co-nanced by the European Commission - DG Environment
(Grant Agreement 070402/2007/460822/SUB/A3).

R EFERENCES
[1] N. Xu, S.Rangwala, K.K. Chintalapudi, D. Ganesan, A. Broad, R. Govindan, and D. Estrin, A wireless sensor network for structural monitoring,
in Proc. ACM SenSys 2004, pp. 1324, Nov. 2004.
[2] J. Paek, O. Gnawali, K. Jang, D. Nishimura, R. Govindan, J. Caffrey,
M. Wahbeh, and S. Masri, A Programmable Wireless Sensing System
for Structural Monitoring, in Proc. IASCM 4WCSCM, pp. 113, Jul
2006.
[3] S.N. Pakzad, G.L. Fenves, S. Kim, and D.E. Culler, Design and
Implementation of Scalable Wireless Sensor Network for Structural
Monitoring, in ASCE Journal of Infrastructure Systems, vol. 14, no.
1, pp. 89101, March 2008.
[4] K. Mechitov, W. Kim, G. Agha, and T. Nagayama, High-Frequency
Distributed Sensing for Structure Monitoring, in Trans. of the Society
of Instrument and Control Engineers, vol. 1, no. 1, pp. 109114, 2006.
[5] S. Kim, S. Pakzad, D. Culler, J. Demmel, G. Fenves, S. Glaser, and
M. Turon, Health Monitoring of Civil Infrastructures Using Wireless
Sensor Networks, in Proc. ACM/IEEE IPSN 2007, pp. 254263, Apr.
2007.
[6] M. Whelan and K. Janoyan, Design of a Robust, High-rate Wireless
Sensor Network for Static and Dynamic Structural Monitoring in Journal
of Intelligent Material Systems and Structures, vol. 20, no. 7, pp. 849-863,
Oct. 2009.
[7] M. Ceriotti, L. Mottola, G.P. Picco, A.L. Murphy, S. Guna, M. Corr,
M. Pozzi, D. Zonta, P. Zanon, Monitoring Heritage Buildings with
Wireless Sensor Networks: The Torre Aquila Deployment, in Proc.
ACM/IEEE IPSN 2009, pp. 277-288, Apr. 2009.
[8] Kistler, ServoK-Beam Accelerometer 8330A datasheet, http://www.
kistler.com/mediaaccess/000-542e-07.06.pdf.
[9] Texas Instruments, TLE207x Low-Noise High-Speed JFET-Input Op
Amps datasheet, http://www.ti.com/lit/gpn/tle2071.

[10] The Arduino project, http://www.arduino.cc/.

[11] Libelium Comunicaciones Distribuidas, SquidBee mote, http://www.
libelium.com/squidbee/.
[12] IEEE, Wireless Medium Access Control (MAC) and Physical Layer
(PHY) specications for Low-Rate Wireless Personal Area Networks
(WPANs), IEEE Std. 802.15.4-2006, pp. 1-305, 2006.
[13] Koan, Koala Nano PC. http://www.koala.it/.
[14] M. Beltrame, Sviluppo di una rete wireless di sensori per il monitoraggio di strutture in tempo reale, Master Thesis, University of
Pavia, Dec. 2008 (in Italian). Available: http://netlab-mn.unipv.it/thesis/
MSc/Beltrame Thesis.pdf
[15] E. Goldoni and P. Gamba, PRISM: a novel protocol for real-time
synchronous acquisitions in WSNs, submitted to ACM Trans. on Sensor
Networks, 10 March 2010.
[16] V. Cionca, T. Newe, and V. Dadarlat, TDMA Protocol Requirements
for Wireless Sensor Networks, in Proc. IARIA SENSORCOMM08, pp.
3035, Aug. 2008
[17] S. Ganeriwal, R. Kumar, and M. B. Srivastava, Timing-sync protocol
for sensor networks, in Proc. ACM SenSys03, pp. 138149, Nov. 2003.
[18] M. Maroti, B. Kusy, G. Simon, and A. Ledeczi, The ooding time
synchronization protocol, in Proc. ACM SenSys04, pp. 3949, Nov.
2004.
[19] O. Mirabella, M. Brischetto, A. Raucea, and P. Sindoni, Dynamic Continuous clock Synchronization for IEEE 802.15.4 based sensor networks,
in Proc. IEEE IECON 2008, pp. 24382444, Nov. 2008.
[20] Fondazione EUCENTRE, European Centre for Earthquake Engineering,
http://www.eucentre.it.
[21] M.A. Hebel, R.F. Tate, and D.G. Watson, Results of Wireless Sensor
Network Transceiver Testing for Agricultural Applications, in Proc.
ASABE Annual International Meeting 2007, paper no. 073077, Jun. 2007.