Sunteți pe pagina 1din 6

An Empirical Analysis of Ad hoc Routing

Protocols for Hybrid Wireless Sensor Networks


M. Saleem and M.A. Sheikh
Centre For Advanced Studies in Engineering (CASE)
Islamabad 44000, Pakistan.
mlsa1eem(jcaseed , mas)case.edu. k
Abstract- This paper investigates and evaluates the possible use
of existing ad hoc routing protocols for hybrid wireless sensor
networks. The key idea is to adopt the routing protocols
designed for Mobile Ad hoc Networks (MANETs) in a hybrid
sensor network that consist of resource enriched nodes (which
may be mobile or static) and resource constrained node s
(static). The role of resource constrained sensor nodes is to

deliver the sensed data to a nearest resource enriched node that


can either store/process the data or transmit it to a central base
station using high power transmission. By virtue of this, we

might be able to save time and effort involved in the design of a


completely new routing protocol from scratch for such a hybrid
and more practical scenario. Our study is inspired by the
similarities found in both MANETs and wireless sensor
networks e.g. ad hoc arrangement of nodes, energy constrained
devices, wireless communications etc. Based on the similarities
of the two networks, we formulated a set of design principles on
the basis of which a protocol designed for ad-hoc networks may
be declared feasible for a hybrids sensor network. Our work is
supported by an experimental analysis in which an example
protocol primarily designed for MANETs has been shown to
perform efficiently in a resource constrained sensor network.

Keywords: Mobile ad hoc networks, wireless sensor networks,


routing protocols, sensor nodes, and energy conservation.
I.

INTRODUCTION

Recent developments in Micro-Electro-Mechanical Systems


and wireless communications have made possible the
existence of a new class of networks, termed as Wireless
Sensor Networks (WSNs), which will revolutionize the
world by redefining our whole lifestyle. Being declared as
one of the most important technologies of the 21St century
[1], WSNs are emerging as an important tool for tracking
contamination in hazardous environments, habitat
monitoring in the nature preserves, enemy tracking in
battlefields, traffic monitoring, surveillance of buildings etc.
WSNs consist of small tiny devices, termed as sensor nodes,
which collect and disseminate sensed attribute to a base
station.
Extensive work has been done in design and development of
specialized routing protocols for wireless sensor networks in
the past few years as surveyed in [2]. But majority of the
work is based on the assumption that all nodes in the
network are homogeneous which is not a viable option.
Instead multi-tiered hierarchical design is more feasible
alternative in which nodes are of different capabilities. We
propose a two-tiered hierarchical network in which static
nodes with least hardware resources perform the sensing task
and communicate the sensed variable to the nodes at second
level of hierarchy. The nodes at the second level may be
mobile and are capable of transferring the received data to

any other base station using long-range transmission, store or

process locally. A practical example of such a network is a


battlefield environment. Mobile soldiers may be outfitted
with powerful data transmitters (resource enriched nodes)

than sensor nodes deployed on ground (resource constrained

nodet ectiveto esimply rort icformato

(e.g.

enemy detection, landmine presence, convoy vehicles) to the


nearest available soldier, who can forward it to other soldiers
or a central base station as shown in Fig. 1.
Resource Enriched Nodes
*
*
4* *

*
*
/ *
* *
Resource Oonstrd Nodes
*
*
*
* *

*
* *

Fig. 1: Two-tiered hybrid sensor network

Due to the ad hoc nature of wireless sensor networks, they

resemble Mobile Ad hoc Networks (MANETs) in many


respects. Topology of both the networks is ad hoc in nature
and energy is an expensive resource that adversely affects

the lifetime of the network. Both networks use wireless


communication, which is inherently error prone. In the
proposed two-tiered hierarchical ad hoc sensor network, we
believe that the similarities of the networks can be exploited
to solve the problem of routing in such a heterogeneous
environment. We propose that some of the existing ad hoc
routing protocols that are optimized for energy efficiency
and low control overhead can be used in the proposed
network architecture. This will not only save a huge effort
required for redesigning a new series of protocols for such a
hybrid network, it will also be a ready to go solution. We
quantified our suggestion by extensive simulation
experiments in which an ad hoc routing protocol originally
designed for MANETs is shown to perform well in wireless
sensor networks. To the best of our knowledge, there is no
such work done in the past and we believe that this work will
open new research dimensions.
Rest of the paper is organized as follows. Section II presents
some common and distinguishing factors of both M\ANETs
and WSNs. Based on similarities and differences in both
categories of the network, we formulate some design criteria
in section III on the basis of which the feasibility of using
existing ad hoc routing protocols in WSNs may be evaluated
qualitatively. Section IV presents a brief description of the
few energy efficient ad hoc routing protocols that may be

adopted in the proposed hybrid network. Experimental


framework and results are discussed in section V. Finally, we
conclude our discussion in section VI.
ITT
MAJOR DESIGN FACTORS OF AD HOC
NETWORKS
Similarities
A.
Network Size: Network size refers to either number of
nodes or geographical area covered by the network. This
parameter is crucial for design of distributed control
mechanisms.
Connectivity: Connectivity refers to two most important
issues. It may refer to the link capacity between two nodes or
it may refer to number of 1-hop neighbors. Links with the
neighbors can be unidirectional or bi-directional. Presence of
unidirectional links can severely affect the overall
performance of protocol and hence should be taken as
critical design parameter. Knowledge of directly connected
nodes is also important and can be exploited in protocol
design.
Network Topology: Topology of ad hoc networks can
change quite frequently due to nodes movement or nodes
failure. One of the vital characteristics of an efficient routing
protocol is to adapt to these rapid changes in network
topology,
Limited Energy Resources: Nodes in ad hoc networks are
battery-powered and hence energy conservation mechanism
must be an integral part of all protocols designed for these
networks. This problem is even more severe in WSNs in
which nodes have non-rechargeable low-power batteries.
Operational Environment: The operational environment
refers to a terrain that prevents Line-Of-Sight (LOS)
operation. This situation is even worse in case of wireless
sensor networks where tiny sensor nodes are usually
scattered randomly in the targeted region. Under these
unfavorable conditions, messages may be lost too often and
as a result the overall operation of the network is disrupted.
Multi-hop Network: A path between a source and a
destination node may involve multiple nodes. This multi-hop
path requires the intermediate nodes to act as routers for
relaying messages to the next hop. An efficient routing
protocol must ensure that routing mechanism is simple
enough so that the nodes may not drain their energy
resources too quickly due to this additional routing activity
resulting in shorter lifetime of the whole network.

limited battery capacity. Majority of potential applications of


WSNs demand the lowest price of a sensor node, as low as 1
dollar, which in turn requires lesser hardware. Hence the
protocols must be targeted for low complexity, low control
overhead and smaller routing tables.
Additional Constraints:
* Due to sheer number of nodes, assigning global identity
to sensor nodes may not possible [2][ 5].
Almost all applications of wireless sensor networks
require the flow of data towards or away from a base
station called sink node. There is absolutely no such
concept involved in MANETs where data flows are
*

irregular.

Routing protocols in WSNs depend upon particular


application. One protocol may be suitable for one
application but not the other one. There is no such
constraint in NIANETs.
III. SELECTION CRITERIA FOR EXISTING ENERGYAWARE AD HOC ROUTING PROTOCOLS

Requirements of WSNs are so much diverse that selection of


a set of protocols suitable for all kinds of applications is
simply not possible. Even the protocols designed specially
for WSNs are application dependent and it cannot be claimed
that a protocol can fulfill the requirements of every
application. A gai in one particular interested parameter
may result in loss of the other important one. Hence, a
compromise must be made to handle the conflicting
requirements. Based on our discussion in the previous
section, we formulated some design principles that must be
fulfilled by an ad hoc routing protocol to become a viable
option in a hybrid network as proposed in the paper. A brief
description of all these desired features is given below.
On-demand Routing
A.
MANETs routing protocols may be proactive or reactive
depending upon the mechanism used to build path between a
source and a destination. In proactive protocols, routes are
determined in advance and stored at each node in the form of
routing tables [3][6][7]. In reactive protocols, routes are
determined only when some data needto be transmitted to a
destination node and hence such protocols are energy
efficient in nature [8]u[9]. The protocols that are hybrid in
nature use combination of both approaches [3][4]. Ondemand routing is preferable in WSNs due to their energy
efficiency, less control overhead and low memory
requirements.

Distinguishingfeatures of Wireless Sensor Networks


No Source Routing
Fault Tolerance: Sensor networks are deployed for B.
extended period of time and operate in an unattended way. Many of the on-demand ad hoc routing protocols use source
Moreover, network topology is usually ad hoc and changes routing mechanism in which intermediate nodes keep adding
continuously due to nodes failure or redeployment of new their addresses to Route Request (RREQ) packets until the
sensor nodes, which may disrupt operation of the whole destination is reached or an intermediate node responds with
network. Fault tolerance is the ability to sustain network Route Reply (RREP) [4]. This approach can severely affect
functionalities without any interruption that must be a the lifetime of the sensor nodes keeping in view that sensor
nodes are deployed in huge numbers. RREQ packet may
characteristic of the protocols designed for WSNs [5].
Scalability: Sensor nodes are deployed in huge number and traverse few hundred or thousands of intermediate sensor
higher densities [5]. So the protocols must be scalable to nodes which will result in extremely long RREQ packet
requiring a lot of energy for transmission. Hence, scalability
tackle such a high population of sensor nodes.
Hardware Limitations: One of the most dominant Of such source routing algorithms is a significant problem.
constraints of sensor nodes iS their limited hardware
resources like low processing power, small memory and
B.

We recommend the use of next hop routing protocols for the


hybrid sensor networks.
C.
Low Control Overhead

makes transition to active state and starts data transmission.


In the listening state, node listens to the messages for TL time
units. The node may switch to active state if it gets a routing
message making it a part of the route or it has some data to
Ad hoc routing algorithms usually discover path via pure send. Otherwise, it makes transition back to the sleeping
flooding or variations of it ignoring the energy consumption state. In the active state, node transmits or relays data and
involved in this process. Therefore, protocols that discover switches back to sleeping state if it is not involved in any
such activity for TA seconds. All state transitions are shown
path at low control overhead are preferable.
in Fig. 2.
D
A dap t ive Fideity/Max im izat io n of Ne tw o rk ife time BECA can be applied as modifications to any ad hoc routing
Sensor networks are expected to be operational for an protocol (e.g., AODV and DSR) to make it energy aware and
extended period of time and hence lengthening the lifetime feasible for the hybrid network. BECA can also work on top
of the network is an important consideration. There are of existing ad hoc routing protocol without modifying the
various approaches adopted by different protocols to achieve underlying routing protocol. However, to be able to work on
this target. Some of them use high-energy nodes to build the top of existing ad hoc routing protocols, we must ensure
routing path [9]. Since sensor nodes are usually deployed in synchronization among the sleeping and active nodes so that
greater densities, protocols that can make use of these RREQ packets may ultimately get through to reach
densely populated nodes to adjust the routing fidelity can destination to establish a path.
also
increase
the
network
lifetime.
For example, in dense networks, some nodes are
sleeping
interchangeable for routing purposes [10]. This information
Tast taffic
send lIca
can be used to extend the lifetime of the network. Hence
protocols that are specially targeted for balanced energy
afer
activ
consumption across the whole network are preferable.
TSafter
no trafi C
L
E.
Ultra-Low Power Consumption
t g
Power conservation is the prime concern since many of the
traffic
nodes in the proposed hybrid network have extremely low
energy resources. Hence, protocol with special mechanism
Fig. 2: States in BECA
for energy saving without disrupting the functionality of the
Max-Min Remaining Energy Routing Protocol
network are the most supported algorithms. Therefore, we do B.
not consider the ad hoc routing protocols (e.g. AODV) that
(MREP)
are not explicitly targeted for low energy consumption.
For a given source S and destination t, remaining energy of
the path P is defined as,
F.
No GPS-Based Algorithms
A t}t
E(P)= min{Ex a xCP
EmainingEn
nod x9
Algorithms that make use of GPS to locate the position of
Ex iS
node
x.
P [9] finds a
at
energy
destination nodes for adapting transmission power to
conserve energy are not feasible for the sensor nodes with Ep*) i cmaximz among alldthseipathconectig Sth
er for
limited hardware resources due to two main reasons. Firstly, pair mRE ifcues the nto tha usesu ni
GPS receivers are themselves too complex and expensive in all transmissions.
terms of power. Secondly, sensor nodes may be deployed In rouemiscovoa
e
rgyteshold
inside the phenomena e.g. building, where GPS receivers of
of energy
threshold Lroand
and fooste
floods theneto
network with
with ananR
RREQ
won't work at all.
packet carrying the value of LI to find a route to destination
IV.
SELECTED ENERGY-AWARE AD HOC ROUTING t. If no RREP is received within certain time period, an
PROTOCOLS
RREQ packet is broadcasted again with new energy
threshold L2 which is less than previous threshold by some
Based on our selection parameters listed in section III, we amount. This relaxation process continues until the path is
have selected few sample energy efficient ad hoc routing found or energy threshold LM (M>1) reduces to zero. In the
protocols that may be used in the proposed hybrid sensor later case, all nodes are obligated to participate in route
network. A brief description of these protocols is presented searching process.
below.
An intermediate node u forwards the non-duplicate RREQ
packet further only if its remaining energy level E,, is equal
A.
Basic Energy Conserving Algorithm (BECA)
to or greater than the threshold that RREQ carries. Before
The fundamental goal of BECA is to minimize energy by forwarding, intermediate node records the node from which
keeping the radio power-off as much of the time as possible it received RREQ packet for backward learning. Intermediate
trading higher latency for reduction in energy consumption node can generate RREP back to source if the path remaining
[10]. BECA defines three states for a node; sleeping, energy E(P(u,t)) is not lower than the energy threshold
listening and active. All nodes start firom sleeping state and carried by incoming RREQ.
do not consume any energy in this mode. After sleeping for Destination node t selects the path with maximal path
Ts time units, node switch to the listening state. However, if remaining energy or the path with the minimum hop count if
the node has some data to send during sleeping state, it it receives more than one RIREQ with same remaining

tWhere

remaisoug
La

MRdn

energy. RREP is then sent back to source node S to notify


successful discovery of such a path. Route maintenance is
similar to AODV in which RERR packet is sent back to
source node.
C.
Mobile Ad hoc on Demand Data Delivery Protocol
(MA ODDP)

the edges through which message travels. Hence, the nodes

not involved in routing process may go to sleep with


probability p1 = (1 - p) where p is gossip probability as
define above. GSP puts a percentage p1 of nodes in sleep

modes where p, is termed as gossip sleep probability. GSP


assumes that the network is synchronized. The length of
period is pre-defined and known as gossip period. GSP
works as given below.
* At the beginning of gossip period, each node decides of
going to sleep mode with probability Pt or stay awake
with probability 1 - pi.
* All sleeping nodes wake up at the end of each gossip
period.
* All nodes repeat this process for every gossip period.

MAODDP [11] is a simplified version of AODV, which


minimizes the control overhead involved in route discovery
and maintenance process in two ways. First, MIAODDP
prohibits the generation of RREP packets by intermediate
nodes. Second, it does not require HELLO packets to verify
the validity of existing routes as in AODV. The protocol
consists of three main phases; route discovery and data
delivery process, routing table management, and local
EVALUATION
conciiy* management.FRMW
aaeet
V. EMPIRICAL AND PERFORMANCE
connectivity
K
Whenever any node has data to send to a particular FRAMEWORK
destination, it generates an RREQ packet along with the data Our empirical evaluation is based on the argument that if the
and broadcasts to all its neighbors. The route request selected protocols work well in static low energy sensor
contains sequence number, broadcast ID along with networks, they may be used in the proposed hybrid model
source/destination addresses. If the source node does not because their validity for relatively high-energy mobile
receive an acknowledgement within certain time period, it nodes is already established. We used Prowler [13], a
re-broadcasts the same route request and the data packet. The probabilistic simulator for wireless sensor networks, for
process continues until a path is found or the number of tries evaluation of MAODDP against AODV. Prowler provides a
expires. After setting up a reverse link to the node from simple but realistic radio and Media Access Control (MAC)
which it received the RREQ packet, an intermediate node models for the Berkeley mote platform. It also supports an
further broadcasts it. When destination node receives this event-based structure similar to TinyOS/NesC that makes it
packet, it sends acknowledgment back to the source nodes. easy to deploy an algorithm on real hardware sensor nodes.
All intermediate nodes setup forward links in response to this We utilized the features of RMIASE [13] framework, which
acknowledgement. No intermediate node is allowed to send is implemented as an application in Prowler.
acknowledgement back to source even if it has a valid route
entry. Nodes maintain a routing table containing the
0.06 FIODV
destination address, sequence number, next hop, expiration
time and active neighbors using that route. If node receives a
0.055 route with an updated sequence number or better metric (less
a,
hops), old entry is replaced with the new one.
0,05
Information about other nodes is obtained in three ways.
First, nodes send HELLO packets when they join the
- 0.045
network. Second, when intermediate nodes receive route
1
0
request and data packets, they update their routing table.
l0
0.04
5
15
20 Nb2e5r 305
40
45
50
Finally, acknowledgement packets sent by destination nodes
are used to obtain this information.
Fig. 3: Latency
This is a highly improved version of AODV and can be very
tested the performance of both protocols, MAODDP and
in the proposed hybrid model. In the results section,
AODV in a static scenario in which source and sink nodes
we evaluated the performance of this protocol against AODV
'
were stationary. The scenario was tested with varying
for static WSNs and observed very encouraging results.
number of nodes; 9, 16, 25, 36 and 49 nodes placed
D.
Gossip-based Sleep Protocol (GSP)
randomly in the sensing field. We assumed symmetric links
nodes. Initial energy level of the nodes was set to 5
GSP is one of those protocols that are generalized for ad hoc between
All ten
Joules (J).from
values are
the values
reported
average of
runs.anAlthough
a number
of
obtained
independent
networks wIth special focus owree ss seso
silike
efrac
erc r
esrd u u opg
BECA. It iS targeted for three main objectives; simplicity,
scalability and connectivity [12]. In ad hoc networks, li.matne reprtc thefollowing:
gossiping is a new idea to reduce the control traffic caused Latency: It' iS defined as the difference in time when an
Ltny ti eie stedfeec ntm hna
by pure flooding. Instead of allowing every node to forward event
iS generated at a source and when it got delivered at the
packets, gossiping permits this forwarding based on a certain siknd.Wreotargelecy
prbbiiy knw sgsi rbbl. Iti.hw n[2 Packet Delivery Ratio: It iS the ratio of total number of
tha gien lageetwrk itha gssi prbablit p reaerevents received at a sink node to the total number of events
than a certain threshold; every node in the network can geradbytesucnosinheesrntwk(n
receive that broadcast message... h
GSP is based on the idea that if gossiping can ensure the event is dispatched ina packet).
delver ofmesageto ll ods i th newor, ten he Energy Consumption: We report total energy consumed by
nodes involved in the forwarding process are connected by thseorndsiantwkdungacexrmn.

efficient~~~~~~~~~

Energy Variance: This parameter indicates the variation of


used energy across the whole network. Lower the value of
this metric; the better is the lifetime of the network.
A.
Results
Fig. 3 shows the latency of both protocols. It is evident that
there is not much difference in the latencies achieved by
AODV as well as MAODDP. The reason is really simple
because both protocols route packets through shortest paths.
In some cases, MAODDP achieves better latency due to its
data delivery along with the RREQ packets.

o '0.9

RODD

350

>

-< 250

L\
50

1
5

10

DeVI.

RODV

0.8-

15

2 2 3
2Number of Nodes

40

45

50

Fig. 6: Energy Variance

CONCLUSIONS

In this paper, we investigated and analyzed the use of


existing energy-efficient ad hoc routing protocols for

0.6u

0.4-

0.5

___________________________________
10 15 20 Number
25 of 30Nodes 35 40 45

the

hybrid sensor network which consists of resource enriched


//(mobile or static) nodes and resource constrained (static)
nodes. The basic function of the resource-constrained nodes
in this network is to sense the physical attribute and
communicate it to a nearby resource enriched sensor node.
Our experimental results clearly demonstrate that the
selected ad hoc protocol (MAODDP) performed efficiently
in low energy sensor networks, which in turn prove its ability
to perform well in the proposed hybrid model. We believe
that our efforts can save the drudgery and time involved in
design and evaluation of entirely a new family of routing
protocols for such hybrid networks. Our future activities will
be focused to evaluate all such protocols in a hybrid network
and compare it with the routing algorithms specifically
designed for wireless sensor networks.

Fig. 4: Packet Delivery Ratio


Fig. 4 represents the packet delivery ratio attained by AODV
and NIAODDP. The success rate of M\AODDP is better than
AODV primarily because its RREQ packets also carry the
data packet. Both protocols suffer in smaller networks due to
pure flooding which makes the network unstable.
Fig. 5 shows the total energy consumed by both protocols
during the whole experiment. It is evident from the figure
that M\AODDP consumes much less energy as compared to
AODV. Better energy consumption is attributed to the fact
that MAODDP reduces the control traffic by suppressing the
REFERENCES
generation of multiplee RREP packets and thee HELLO
"21 Ideas for the 21st Century," Business Week, pp. 78-167, Aug.
[1].
vldt Of liks. nce energy iS
i
packets for checking the validity
30, 1999.
very scarce resource in both types of ad hoc networks, this [2].
J.N. Al-Karaki, A.E. Kamal, "Routing techniques in wireless sensor
networks: A survey" IEEE Wireless Communications, 11(6):6-28,
will increase the network lifetime.
Dec. 2004.
Energy variance is shown in Fig. 6, which depicts that
Elizabeth M. Royer,
Santa Barbara,
Chai-Keong Toh, "A Review of
in smaller topologies is very poor. [3].
energy balancing
balancing
for Ad Hoc Mobile Wireless Networks,"
Routing Protocols
Howenergy it showsbetterbehavioronlargertopologieCurrent
shows better behavior on larger topologies,
IEEE Personal Communications, April 1999.
However,
which is desired for large-scale networks. Higher energy [4]. C.E. Perkins (Editor), "Ad Hoc Networking," Addison-Wesley,
2001.
variance is also because of the fact that the protocols use

generationfo chekin

usingle

olikes Snd

shortest path. This variable

can

HEnr

be improved by Simply

energymetric.Erdal Cayirci

[5].

using an altemate
alternate energy metric.

[6].

140

-50

120

-0-08- HADOUP

RODV

E 0-

s0

a,

4
40I-<
L
20

\
\

F-

a5

[7].

10

15

l
20

25

30

Number of Nodes

35

Fig. 5: Total Energy Consumed

40

45

50

Ian F. Akyildiz, Weilian Su, Yogesh Sankarasubramaniam, and


"A Survey on Sensor Networks," IEEE

Communications Magazine, August 2002.

C. Perkins, P. Bhagwat, "Highly Dynamic Destination-Sequenced


Distance-Vector Routing (DSDV) for Mobile Computers," Proc. of
ACM SIGCOMM Conference on Communications Architectures,
Protocols and Applications, pp.234-244 Oct. 1994.

S.

Murthy, J. J. Garcia-Luna-Aceves, "An Efficient Routing

Protocol for Wireless Networks," ACM Mobile Networks and App.


J, Special Issue on Routing in Mobile Communication Networks,
Oct. 1996, pp. 183-97.
C. E. Perkins, E. M. Royer, "Ad hoc On-Demand Distance Vector
[8].
Routing," Proc. 2nd IEEE Wksp. Mobile Comp. Sys. and Apps.,
Feb. 1999, pp. 90-100.
Baoxian Zhang, Hussein T. Mouftah, "Adaptive Energy-Aware
~~~~~~~~~~~~~~[9].
-Routing Protocols for Wireless Ad hoc Networks," Proceedings of
the First International Conference on Quality of Service in
~~~~~~~~~Heterogeneous Wired/Wireless Networks (QSHINE'04), 2004.
[10]. Ya Xu, John Heidemann, Deborah Estrin, "Adaptive EnergyConserving Routing for Multihop Ad hoc Networks," USC/ISI
research report 527, October 12, 2000.
[11]. Humayun Bakht, Madjid Merabti, Bob Askwith "A Routing
Protocol for Mobile Ad-Hoc Network," In the proceeding of 1st
International Computer Engineering Conf, December 27-30 2004,
Cairo, EGYPT.

[12].
[13].

Z. J. Haas, J. Y. Halpern, L. Li, "Gossip-based Ad hoc Routing," in


Proceedings of IEEE INFOCOM, 2002.
Y. Zhang. Routing modeling application simulation environment.
http://www2.parc.com/isl/groups/era/nest/Rmase/.

S-ar putea să vă placă și