16th Mediterranean Conference on Control and Automation

Congress Centre, Ajaccio, France

June 25-27, 2008

Packet Error Rate Optimization for Routing algorithms in ad hoc

wireless sensor networks
Jordi Sabater, Manel Lpez, Jos M. Gmez, Atil Herms
{jsabater, mlopez, jmgomez, herms}@el.ub.es

Electronics Department, University of Barcelona, c/. Mart i Franqus 1, planta 2, 08028 Barcelona

AbstractOver the last years, wireless multihop ad hoc

networks have received a tremendous interest from research
groups. The major focal point has been the routing protocols,
i.e. how to select the best route to send information from any
source to any destination. Each node in an ad hoc network may
work as a router to relay connections or data packets to their
destination. The key issues of ad hoc networking are Medium
Access Control, which is used to share common channel
resources among wireless nodes, and the networking layer.
In this work, we have implemented a routing protocol based on
the multihop algorithm to transmit information from any
source to any destination node. We have focused on an indoor
wireless sensor network, where all the links between nodes
follow a line-of-sight propagation model. Medium Access
Control and Physical layer have been implemented according to
the IEEE 802.15.4 standard in the 2.45 GHz frequency band
where information is modulated in O-QPSK.
In multihop wireless networks, one important question is how
does the route selection depend on the physical and logical
topology of the network. In terms of routing there is a trade-off
between between hop distance and reception rate. This
compromise will give us the best route to minimize the power
consumption and maximize the transmission success
probability.
Index Terms Routing algorithms, Wireless Sensor
Networks, IEEE 802.15.4 standard, software algorithms

I.INTRODUCTION

D HOC networking is an increasingly important topic

in wireless communications and has been regarded as
one of the key features of wireless sensor networks. A sensor
network consists of a large number of sensor nodes that are
placed inside or near the device or phenomenon under test
[1,2]. Those nodes collaborate to measure and to sense
parameters and deliver the information over the network.
Each node in an ad hoc network can work as a router, to
relay connections or data packets to other nodes until it
arrives to the wished destination node. Due to the fact that
this kind of nodes are usually powered by batteries with a
limited lifetime, one of the major concerns is related to the
power consumption [3]. Then it is clear that the main issue
of the performance and viability of such systems relies
strongly on the ability of the network to save energy on each

978-1-4244-2505-1/08/$20.00 2008 IEEE

node. Moreover, the information collected by the nodes

should be processed to avoid redundancies and minimize the
energy spent during the transmission [4,5].
Among the different factors that can be considered while
designing a routing protocol for wireless communications
are energy efficiency, delivery latency, packet success
probability, adaptability and scalability [6]. All these
parameters are used by routing algorithms to determine route
optimality, usually it is refereed as the route-metric or simply
the metric. Most routing schemes try to find shortest paths in
terms of hop count [7-9]. Other algorithms consider not only
the minimum number of hops but also the link quality and
the energy of the relayed nodes for this transmission. It is
clear that a good routing algorithm must be energy efficient.
Intermediate nodes are also often involved in route
filtering and selection. This is clear because route filtering
and selection is based on local cost and thus does not
generate globally optimized routes. Although most routing
protocols only handle a single path, some others provide
mechanisms to build and maintain multiple paths between
two communication peers.
For ad hoc sensor networks, at network layer level, one
important question strongly related to the power
consumption is whether it is more advantageous to route
over many short hops (short hop routing, and usually taken
to the extreme nearest neighbor routing) or over a smaller
number of longer hops. One of the key issues to choose
between short hop and long hop routing is the interference.
According to Ephremides et al. [10], when a larger number
of short hops are replaced by a smaller number of long hops
It is unclear whether more interference is caused by a single
transmission at higher power or multiple transmissions at
lower power. Indeed, a shorter transmission at higher power
may permit more efficient reuse of the communication
channel.
Haenggi et al[11], are opposed to this argument and
remember that the Signal to Interference Rate (SIR) does
not depend on absolute power levels. Thus, increasing the
transmit power levels in the network by the same factor does
not have a negative impact on any packet reception

1839

probability in the network. Haenggi concludes that long

hop transmission does not inherently cause more
interference.
Noise is the other key factor that contributes to the loss of
packets during the retransmission. The well known Signalto-noise-and-interference rate (SINR) plays then an
important role in the total energy of the network. Any packet
loss implies the retransmission of that packet, and the spent
of energy in all those nodes entrusted to retransmit the
information.
In this paper we will focus in the determination of the best
route depending on the number of hops and also the
visibility of the nodes. The power transmission will be
regulated depending on the distance among sensors, the
power consumption and the success probability threshold.
We will discuss the evolution of the transmitted packets as a
function of the propagation loss, the antenna loss, the
minimum SINR condition and the stochastic nature of the
wireless channel. In our case, there are a sufficiently large
number of paths, due to the reflections of walls, ceiling, and
furniture, to consider that the channel can be modeled as a
Rayleigh fading channel [12] for NLOS channels and also as
a Rician fading channel for LOS channels.
For those kind of channels, some authors [13-15] reported
that even with static nodes, as we are considering in this
paper, the channel quality varies because any movement
environment affects the multipath geometry of the RF signal.
The significant variation of the link quality when nodes are
fixed is also pointed out in [16].

applications in the 2.4 GHz unlicensed ISM band. This RF

transceiver is compliant with the IEEE 802.15.4 standard.

II. EXPERIMENTAL

(b)

The sensor nodes used in this work are the well known
Tmote sky (usually known as motes), a compliant device for
IEEE 802.15.4 wireless communications [17]. The
microcontroller used is a 8MHz Texas-Instruments 16-bit
MSP430 with 10K of RAM and 48KBytes of Flash memory
[18]. Tmote sky includes a real time embedded operating
system: TinyOS, which is specifically designed for wireless
sensor networks. Its component-based architecture permits
the abstraction between hardware and software. It contains a
minimal set of code, due to the severe memory constraints of
the Tmote sky node. Otherwise, its component library
includes network protocols, distributed services, sensor
drivers and data acquisition tools [19, 20].
The sensing unit contains a Sensirion SHT15, it is a
temperature and humidity digital sensor. It also contains a
Hamamatsu S1087-01 photodiode for sensing the entire
visible spectrum and near infrared [21]. The sensing
information is packed in an 802.15.4 frame format and sent
to the medium from the communication facility.
The communication facility is based on a CC2420 from
Chipcon [22]. The CC2420 is a low-cost transceiver
specifically designed for low-power, low-voltage RF

For our experimental results, we use two 1.5V batteries

serial connected. The current consumed to transmit and
receive a typical frame of around 40 bytes (mean frame size
for our experiment) is 20.85 mA. Consequently, since the
voltage supply is 3V, the power consumed during the
transmission is around 60 mW. Moreover, from our
experimental results, we detect that the power consumption
for transmission and reception is the same for the particular
case of CC2420 transceiver.
Experimentally, we observe that the consumption for
transmission and reception can shift from the mean value of
60 mW by a maximum of 1mW for these kinds of frames.
Table I (a) shows the current consumption of one of the
nodes that constitute the WSN following the different parts
described in the previous section. Table I (b) shows the
power consumption of the same node for the three operation
modes.
controller

Communication Unit

Sensing Unit

60 W

60.85 mW

15 mW(*)

(a)
Transmitting (0dBm) /
Receiving

Mute

Sleeping

60 mW

2 mW

>60W(**)

Table I. (a) Average power consumption of the different working subparts.

(b) Power consumption depending on the different operation modes. (*)
current per sensor.(**) Power consumption per sensor.

III. DEPENDENCE OF THE PACKET ERROR RATE WITH THE

INTERFERENCE AND NOISE (POINT TO POINT LINK QUALITY)

Point to point link quality is of a great importance for a

successful wireless communication. Depending on the
number of loss packets detected in every hop (that will be
related with the Packet Error Rate) it is possible to choose
the optimal neighborhood node for every individual hop. We
must also take into consideration the power consumption in
every hop, and we will use all this information to choose the
best next hop [11].
As a first approximation to the problem, Packet Reception
Rate has been calculated as a function of the distance for a
uniformly random distribution of wireless nodes. The
experiment is based on the well known disk model [2426], where it is assumed that the radius for a successful
transmission of a packet has a fixed and deterministic value,
irrespective of the condition and realization of the wireless

1840

channel. The mean inter-node distance selected for the

experiment varies from 5 meters to 30 meters. The node
distribution is the same for all the routing algorithms and the
relay node chosen for any retransmission is that given by the
nearest-but-inside the radius line. It is important to remark
that in this experiment, the power transmission was constant
and equal to 0dBm for every distance.
An increase of the distance will imply a decrease in the
SNR. On the contrary, an increase of the number of
retransmissions will imply an increase of the interference. In
this case, there is not a dependence with the power
transmission.
The reception node was located at a fixed distance for
every inter-node distance (40 meters). The shorter is the
mean inter-node distance the larger is the number of
intermediate nodes. The result suggests that, for this
particular case, there is a minimum PER located around 10
meters. This result is represented at Figure 1.

Rate (%)
Packet Reception

Packet Error Rate (%)

15

10

10

Figure 2 Packet Success Rate as a function of the hop distance for a 0dBm
(diamond dot) and -5dBm transmitted power (square dot)

100
80
60
40
20
0

Dis 20 40
tan
ce
(m

60

e te

80

rs)

10
100

15

d
No

20

eD

25

an
ist

30

ce

(m

0
0

determined threshold, the transmitter output power could be

decreased leading to energy saving. On the other way, lower
power transmissions produce less interferences and help to
the achievement of overall network higher throughput.

15

20

25

Radius route distance (meters)

Figure 1 Evolution of the Packet error rate as a function of the mean
internode distance. The inset shows the Packet Reception Rate as a
function of the distance hop. Triangle show a 5 meters relay node
distance, squares 7 meters, stars 10 meters, circles corresponds to a 20
meters, and finally, pentagon corresponds to a 30 meters relay node
distance.

The inset of Figure 1 shows the evolution of the Packet

Reception Rate as a function of the distance for different
radius: 5, 7, 10, 20 and 30 meters hop distance. It is clear
that, in this experiment, exist an optimum distance that
minimizes the packet loss. This hop internode distance will
depend on the transmit power (constant and equal to 0 dBm
in our case), the interference, the signal-noise ratio, and of
course the probability of correct reception.
However in some cases, it makes sense to adjust the RF
power to achieve a good PSR (Packet Success Rate) while
consuming minimum energy. In case that for a certain link
formed by a pair of nodes the PSR would be above a

Figure 2 shows a comparison of overall PSR for a certain

link using 5, 10, 15, 20, 25 and 30 meter LOS hops and two
different RF output power (0 dBm and -5 dBm). The results
have been obtained from simulations that takes into account
a Ricean channel and the work conditions imposed by the
IEEE 802.15.4. As can be observed in Figure 2, the PSR
simulation at 0 dBm (diamond dots) fits quite well with the
experimental result presented at Figure 1.
Other important point to remark is the evolution at -5dBm
(square dots). The decrease of the transmission power can be
directly linked to a decrease of the interference among
nodes and a increase of the SNR. Then this implies a shift
of the maximum PSR to lower mean inter-distance
On the other way, for our particular case, where the IEEE
802.15.4 Standard has been programed at the wireless
sensor node working at 2.45GHz, 60mW of DC are needed
to generate 1mW of RF. That makes system efficiency poor
due to hardware leaks, but it is still interesting to see that our
ideas are still valid.
IV. DEPENDENCE OF THE PER OPTIMITAZION WITH THE
ROUTING ALGORITHMS

Routing algorithms provides a global vision of the

network. The process required to establish a routing table
includes the following steps: (i) Route discovery, (ii) Route
selection, and (iii) Route maintenance and data forwarding.
Figure 3 shows a mesh based wireless network with a
random node distribution. No hierarchic distribution is
considered here, so, every node can send and/or receive
messages to/from other node depending on its routing table
and the destination node. The routing algorithm is based on
the three steps commented previously. At the following
paragraphs these steps are explained in detail.

1841

Figure 3 Random node distribution on a 100x50x40 m 3D Box. The

figure also shows the XY, ZY and XZ projection of the 3D box.

Route discovering and route selection

The sense of route discovery is to find nodes in the
network and then to know potential routes toward desired
destinations.
Every node send n small packets, usually known as
hello packet, in broadcast mode to know the
neighborhood. Those nodes who receive this packet send to
the source the confirmation of reception as well as
information about their vicinity, known nodes, low power
scheduling and others. m packets arrive to the initial node
(n>m). In case that the probability of error (PER) would be
higher than a threshold value, the connection will be
rejected. This algorithm determine which nodes has a better
SNR and those nodes are included at the routing table.
Table II shows the discovering neighborhood and the
error probability of the link. The threshold PER was 30%
At this point, it is very important the cooperation among
nodes since they have to take into account the optimization
of many parameters for the whole link and not only for the
single hops. It can be seen comparing table II (i) and (ii). At
Table II (i) we observe that node 7 is unreachable. In fact,
node 1 can arrive to node 7, but the PER necessary is over
the PER threshold selected for this network (0.3). On the
other hand, cooperation among the different nodes shows
that the whole PER is decreased if we take into account
node 7 as a next hop to arrive to some destination nodes. It
demonstrate that the best node in a single hop does not
necessarily leads implies the best route to reach the
destination node.

Table II. (i)Neighborhood discovering state for node 1. Only

those nodes with a low PER are considered as neighbors. (ii)
Final routing table for node 1

Data transmission and Route maintenance

Data transmission means to forward data packets based on
route information. In our case, the selection of the route is
based on the routing table established and explained in the
above paragraph. Route maintenance refers to the means of
keeping the current route valid and repair it if the route is
not allowed or valid anymore. Routing maintenance in
general implies to refresh the routing table in order to know
if any node has fallen down or it is overloaded. In this case,
it is necessary to start the route selection algorithm again.
From the experimental results presented it seems clear
that the routing protocol definition will depend not only on
the amount of neighbor nodes that exist around the source
but also the link quality. If the source node can establish a
relationship between neighbor nodes and link quality for
every individual neighbor, and cooperation among nodes
exist in order to minimize not only the link PER but also the
whole PER of the route, the routing table can minimize the
PER for all defined routes. Figure 4 describes the route
discovering algorithm as it was implemented experimentally.
As was explained above, the source node have to send n
packets in broadcast mode to its neighbors. In this packet
there is information related to the unknown route, its actual
routing table and the PER associated to this route. The
neighbor nodes that receive these frames update their
routing table with the information sent if necessary, and look
for the destination node in their local tables. After that, those
neighbors whose include the destination node in their

1842

VI. ACKNOWLEDGMENTS
This work has been partially funded by the Spanish
Program of Technology transfer under project num. PET
2005-0019-02, ESP2007-60286 and FIT -330101-2007-4.
[1]
[2]

[3]
[4]
[5]
[6]

[7]
[8]
[9]
[10]
[11]

Figure 4 Chronological evolution of the algorithm implemented to route

data packets minimizing the Packet Error Rate and the energy
consumption of the network.

routing table will respond to the request packet of the source

node, indicating the number of packets received, i.e. m,
(where m n) and including their routing table as a payload.
An upper layer software will determine the best route for
every destination taking into account the global PER. With
this information, the source node can update its routing table
selecting which node is the next hop for any destination one,
i.e. which node minimize the packet error rate to arrive to a
destination, and saving one more possibility of routing path
in order to avoid problems of network saturation.
V. CONCLUSION
This paper describes the PER behavior depending on the
point to point link quality and the optimization that can be
achieved including a global point of view of the PER in the
routing algorithm. We have focused on an indoor LOS
network because of this case is one of the most implemented
examples of wireless sensor networks. We have observed
that a better local signal-noise ratio does not imply a better
source -to-destination route. The selection of a intermediate
node which has a PER above the threshold, can be necessary
to achieve the best source-to-destination route.

[12]
[13]
[14]
[15]

[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]

[25]
[26]

1843

VI. REFERENCES
I. F. Akyldiz, W. Su, Y. Sankarasubramaniam and E. Cayirci,
Wireless sensor networks: a survey Computer Networks, vol. 38,
no. 4, pp. 393-422, 2002
J. L. Burbank, P. F. Chimento, B. K. Haberman and W. T. Kasch,
Key challenges of military tactical networking and the elusive
promise of MANET technology, IEEE Communications Magazine,
vol. 44, no. 11, pp 39, 2006
D. Culler, D. Estrim and M. Srivastava, Overview of sensor
networks, IEEE Computer Society, vol. 37, no. 8, pp 41-49, 2004
J. F. Chamberland and V. V. Veeravalli, The art of sleeping in
wireless sensing systems, IEEE Workshop on Statistical Signal
Processing, pp 17-20, 2003
M. Lopez, A. Osorio, J. M. Gomez, A. Moragrega and A. Herms,
Energy saving strategies for 802.15.4 based greles
communications Proceedings on DCIS 2006
M. J. Lee, J. Zheng, X. Hu, H. Juan, C. Zhu, Y. Liu, J. S. Yoon and T.
N. Saadawi, A new taxionomy of routing algorithms for wireless
mobile ad hoc networks: The component approach, Topics in ad
hoc networks, IEEE Communications Magazine, vol.44, 2006
Request for Comments RFC 1058, Routing Information Protocol.
Request for Comments RFC 2453, Routing Information Protocol v2.
Request for Comments RFC 3561, Ad hoc On-Demand Distance
Vector (AODV) Routing protocol, 2003
A. Ephremides, Energy concerns in wireless networks, IEEE
Wireless Communications, vol. 9, pp 48-59, 2002
M. Haenggi and D. Puccinelli, Routing in ad hoc networks: A case
for long hops, IEEE Communications Magazine, vol. 43, 2005
J. G. Proakis, Digital CommunicationsEd. Mc Graw Hill, 2001
IEEE Transactions on Veh. Tech. Vol. VT-37, 1988. Special issue on
radio propagation
A. Ephremides Energy concerns in wireless networks, IEEE
Wireless Communications, vol. 9, pp 48-59, 2002
A. Woo, T. Tong, D. Culler, Taming the underlying challenges of
reliable multihop routing in sensor networks Proceedings of the
First International Conference on Embedded Networked Sensor
Systems. Los Angeles, 2003
A. J. Goldsmith and S. B. Wicker, Design challenges for energyconstrained ad hoc wireless networks, wireless communications,
vol. 9, no. 4, pp 8-27, 2002
http://www.moteiv.com
http://www.ti.com
http://www.tinyos.net
P. Levis, N. Lee, M. Welsh and D. Culler, TOSSIM: Accurate and
Sclable Simulation of Entire TinyOS applications,SenSys 2003
SHT15 Datasheet, available in http://www.sensirion.com
CC2420 Datasheet, available in http://www.chipcon.com
X. Liu, M. Haenggi, The impact of the topology on the throughput
of interference-limited sensor networks with Rayleigh fading,
proceedings, IEEE SECOND 2005.
R. D-Souza, D. Galvin, C. Moore and D. Randal, Global
connectivity from local geometric constraints for sensor networks
with various wireless footprints, Information Proceedings in Sensor
Networks, IPSN 2006, pp 19-6.
R. Wattenhofer, L. Li, P. Bahl and Y. Wang. Distributed topology
control for power efficient operation in multihop wireless ad hoc
networks. In Proceedings of IEEE INFOCOM, 2001
L Li, J. Y. Halpern, P. Bahl, Y. Wang and R. Wattenhofer. Analysis
of a cone=based topology control algorithm for wireless multihop
networks. Proceedings of ACM symposium on Principles of
Distributed Computing, 2001