Autonomic Wireless Sensor Networks

Shah Sheetal
Viterbi School of Engineering, University of Southern California,
Los Angeles, CA, USA
E-mail: sheetals at usc.edu
Abstract
Wireless ad hoc networks of sensor nodes are envisioned to be deployed in the physical
environment to monitor a wide variety of real-world phenomena. Wireless sensor networks
(WSNs) are becoming popular in military and civilian applications such as surveillance
monitoring disaster recovery home automation and many others. !lmost any sensor network
application re"uires some form of self-configuration and autonomic functionality. #ollowing
$%&s initiatives towards !utonomic computing many architectures and protocols for network
self-organi'ation and management have been proposed and being implemented.
(he paper presents concept of !utonomic )omputing with respect to Wireless Sensor Network.
(he paper introduces Wireless sensor network basics design goals and challenges along with
current and future applications. $t articulates basic needs of incorporating autonomic computing
principles into the design of Wireless Sensor Networks. (he paper also outlines recent
contributions to !utonomic network architectures research pro*ects proposed architectures and
routing protocols for !utonomic Wireless Sensor Networks.
1 Introduction
ireless sensor net!or"s have critical a##lications in the scientific, medical, commercial, and
military domains. E$am#les of these a##lications include environmental monitoring, smart
homes and offices, surveillance, and intelligent trans#ortation systems. %t also has significant
usages in biomedical field. As social reliance on !ireless sensor net!or" technology increases,
!e can e$#ect the si&e and com#le$ity of individual net!or"s as !ell as the number of net!or"s
to increase dramatically.
ireless sensor net!or"s are ty#ically used in highly dynamic, and hostile environments !ith
no human e$istence 'unli"e conventional data net!or"s(, and therefore, they must be tolerant to
the failure and loss of connectivity of individual nodes. )he sensor nodes should be intelligent to
recover from failures !ith minimum human involvement. *et!or"s should su##ort #rocess of
autonomous formation of connectivity, addressing, and routing structures. +ecent researches on
Autonomic *et!or"ing can serve as basis for design of !utonomic Wireless Sensor Networks.
)he #a#er introduces Autonomic com#uting and !ireless sensor net!or" conce#ts. ,iscusses
ho! the fundamental #ro#erties of Autonomic com#uting com#ly !ith the basic design
re-uirements for !ireless sensor net!or"s. .ro#osed #rotocols for ireless Sensor *et!or" and
their a##licability and suitability to Autonomic ireless Sensor *et!or"s and re-uired
im#rovements. )he #a#er gives brief overvie! of research #ro/ects and architectures for
autonomic communication and net!or"ing !hich can be a##lied to S*s. )he last section
focuses on the current and #ossible future a##lications of Autonomic ireless Sensor *et!or"s.
2 Autonomic Computing
2.1 Background
)he dramatic increase in com#uting devices, increased com#uting ca#acity and com#le$ity
combined !ith #o#ularity of internet resulted in #henomenal gro!th in heterogeneous net!or"s
and net!or" a##lications. ith this increasing system com#le$ity, net!or" management issues
and communication #rotocols are reaching a level beyond human ability to manage and secure so
the stability of current infrastructure, systems, and data is at an increasingly greater ris" to suffer
outages and general disre#air. 0uture net!or" algorithms need to be ada#tive, robust, and
scalable !ith fully distributed and self-organi&ing architectures. Automation, self-#rotection and
self management of !ide s#read net!or"s may solve the #roblem till some e$tent.
As the conce#t of self management rooted u#, the most direct ins#iration one can thin" of !as
the autonomic function of the human central nervous system !here autonomic controls use
motor neurons to send indirect messages to organs at a sub-conscious level. )hese messages
regulate tem#erature, breathing, and heart rate !ithout conscious thought. 1bservation and
analysis of these com#le$ ada#tive systems found in nature became a ma/or source of ins#iration
to design algorithms for self-managed, self-organi&ed, self-configuring and self-#rotecting
systems.
)a"ing ins#iration from autonomic nervous system of the human body %23 created a
foundation for autonomic systems by ta"ing initiatives to!ards Autonomic Com#uting for
relieving humans from the burden of managing com#uter systems !hich is gro!ing enormously
till the e$tent of unmanageability. 4567
2.2 Autonomic System
Autonomic System is a system !hich !or"s inde#endently on #redefined #olicies and rules
!ithout any human interaction and manage and configure itself on its o!n based on #redefined
rules and gained "no!ledgs over the time. %23 has defined the follo!ing four functional areas
for self management of Autonomic System: '+ef 4587(
Self-)onfiguration+ Automatic configuration of com#onents.
Self-,ealing+ Automatic discovery, and correction of faults.
Self--ptimi'ation+ Automatic monitoring and control of resources to ensure the o#timal
functioning !ith res#ect to the defined re-uirements.
Self-.rotection+ .roactive identification and #rotection from arbitrary attac"s.
2. IB! Autonomic Computing arc"itecture
%23 Autonomic Com#uting Architecture 4597 defines an abstract information frame!or" for
self-managing %) systems. %n the information frame!or", an autonomic system is a collection of
autonomic elements. Each autonomic element consists of an autonomic manager 'A3( and the
managed resource '3+(. )he communication bet!een the A3 and the 3+ is done through the
3+:s management interfaces, !hich e$#oses t!o ty#es of hoo"s, sensors and effectors. )he
sensors are used by the A3 to obtain the internal state of the 3+, and the effectors are used by
the A3 to change the behavior of the 3+. )he A3 enables self-management of the resource
using a ;;monitoring, analysis, #lanning, and e$ecution:: control loo#, !ith su##orting
"no!ledge of the com#uting environment, management #olicies, and some other related
considerations.
0igure 6. 2asic Autonomic Com#uting +eference Architecture
')his is figure is ta"en from %23 Autonomic Com#uting architectural 2lue .rint 4597(
)he autonomic com#uting information model only #rovides the conce#tual guidance on
designing self-managed systems< in #ractice, the information model needs to be ma##ed to an
im#lementable management and control architecture for Autonomic *et!or"s. S#ecifically,
measurement techni-ues, rule engines, #lanning methodologies, dynamic resource allocation
techni-ues, security and management schemes need to be develo#ed for autonomic elements, and
a scalable management #latform is re-uired to coordinate the autonomic elements into a self-
managing system.
Wireless Sensor Network #
A !ireless sensor net!or" 'S*( is a net!or" that is made of hundreds or thousands of sensor
nodes !hich are densely de#loyed in an unattended environment !ith the ca#abilities of sensing,
!ireless communications and com#utations 'i.e. collecting and disseminating environmental
data(. )hese s#atially distributed autonomous devices coo#eratively monitor #hysical and
environmental conditions, such as tem#erature, sound, vibration, #ressure, motion or #ollutants,
at different locations. )he basic archetecture of ireless sensor *et!or" is sho!n in 0igure9.
0igure 9. 2asic Architecture 1f ireless Sensor *et!or". '+ef 45=7(
Each autonomic node in a sensor net!or" is ty#ically e-ui##ed !ith a radio transceiver or other
!ireless communications device, a #rocessing unit !hich can be a small micro-controller,
sensing unit, and an energy source, usually an al"aline battery. Sometimes, a mobili&er is needed
to move sensor node from current #osition and carry out the assigned tas"s. Since the sensor may
be mobile, the base station may re-uire accurate location of the node !hich is done by location
finding system. )he si&e of a single sensor node can vary from shoebo$-si&ed nodes do!n to
devices the si&e of grain of dust. 45=7
0igure 8. Com#onents of a Sensor *ode '+ef 45=7(
.1 $e%uirements and &esign 'actors in Wireless Sensor Network
0ollo!ing are some of the basic re-uirements and design factors of !ireless sensor net!or"
!hich serve as guidelines for develo#ment of #rotocols and algorithms for S* communication
architecture.
/. #ault (olerance !daptability and 0eliability+ Sensor net!or"s are re-uired to o#erate
through ada#ting to the environmental changes that sensors monitor. )he net!or"s should be
self-learning. +eliability is the ability to maintain the sensor net!or" functionalities !ithout any
interru#tion due to sensor node failure. Sensor node may fail due to lac" of energy, #hysical
damage, communications #roblem, inactivity, or environmental interference. )he net!or" should
be able to detect failure of a node and organi'e itself, reconfigure and recover from node failures
!ithout loosing any information. 45>7
1. .ower )onsumption and .ower management: 1ne of the com#onents of sensor nodes is the
#o!er source !hich can be a battery. )he !ireless sensor node being a microelectronic device,
can only be e-ui##ed !ith a limited #o!er source 45=7. 1ver the remote inaccessible #lace !ith
less human control and e$istence, #o!er sources #lay critical role in survival of sensor nodes.
.o!er source should be intelligently divided over sensing, com#utation, and communications
#hases as #er re-uirement. Sensors can be hibernated !hen inactive. Lots of current researches
are focusing on designing #o!er-a!are #rotocols and algorithms for !ireless sensor net!or"s.
+ecently, solar energy is also considered as an o#tion for em#o!ering remote sensor nodes
!hich are e$#osed environment.
2. Network 3fficiency and 4ata !ggregation: 0looding ra! sensed data over the net!or" can
easily congest the net!or". Some critical a##lications li"e intruder detectors re-uire urgent
transmission and faster #rocessing of data !hich may degrade #erformance and loose reliability
due to congestion or latency in the net!or". %ntelligent aggregation of sensed data and
elimination of un!anted and redundant information and data com#ression can be a solution for
efficient resource and energy utili&ation and congestion avoidance. 3any algorithms li"e
directed diffusion 45?7 are #ro#osed to facilitate data aggregation and dissemination !ithin the
conte$t of S*s.
5. $ntelligent 0outing: %n many a##lications, sensor nodes are moving nodes and can change
#lace dynamically. +outing #rotocols must be ada#tive to these changes and should be self-
healing and self-configuring. )he information should be #ersistent in s#ite of changes in net!or"
nodes. Lo! #rocessing ca#acity of a node creates many challenges for routing #ac"ets
throughout the neighbouring nodes intelligently. As discussed above, some a##lications may
re-uire a faster communication and instant res#onse. +outing algorithms should be intelligent to
choose minimum ho# and minimum distance #aths for data transfer. 45@7
6. &anagement challenge A 3anaging the communication over heterogeneous net!or"s is basic
challenge in self-managed system because #olicies and communication #rotocols #lan an
im#ortant role in net!or" communication. Also, it is necessary to balance the level of detail the
net!or" is #roviding to the client against the rate at !hich energy is being consumed !hile
gathering the data. Clearly, it is #referable to have the net!or" automatically do this tuning,
rather than re-uiring manual intervention.
)hese basic re-uirements and design goals serve as challenge for current technology. )hough
current %. routing #rotocol e$ist and have significant a##lications in current net!or"s and
%nternet, they do not satisfy com#lete design re-uirements in ireless sensor net!or"s because
S* nodes ty#ically has limited com#uting ca#acities and less #o!er. So S*:s re-uire a
different infrastructure and #rotocol stac" !hich can be im#lemented using autonomic
com#uting conce#t as !e !ill discuss in ne$t section.
(. Wireless Sensor Networks and Autonomic Computing #
)o clarify the contribution that autonomic com#uting can bring to ireless Sensor *et!or"s
'S*(, let:s e$amine ho! S* design re-uirements and o#erations can be tac"led using
autonomic #rinci#les.
As discussed above, there can be sensor nodes !hich are moving and can change their #osition
dynamically or even leave the net!or" coverage area. )herefore, a #re-#rogrammed
configuration for the net!or" !ill not !or". Sel'#con'iguring nodes can set u# net!or"
connections, evaluate if there are any ga#s in the S* and re#lace a moved or dead node in the
net!or". Since sensors can be de#loyed in an unattended area 'e.g., forest and ocean( or
#hysically unreachable area 'e.g., inside a building !all(, they are re-uired to o#erate !ith the
minimum aid from base stations or human administrators. Although ma/ority of current sensor
a##lication have already considered this in their net!or" design, there is still a need for S* to
have the ability to reconfigure and recover itself !ithout too much human intervene, es#ecially
in inaccessible environment. 45=, 5>7
Sensor reading usually contains some noises< it may be a false #ositive due to malfunction of
sensors. Sensors are re-uired to collectively sel'#"eal 'i.e., detect and eliminate( false #ositives
in their sensor readings instead of transmitting them to base stations. )his can also reduce #o!er
consum#tion of sensors because data #rocessing !ithin the sensor incurs much less #o!er
consum#tion than data transmission does 4657.
Sensor nodes are generally e$#osed to much harsher conditions than standard com#uting
e-ui#ment, and are thus sub/ect to energy de#letion and incidental damage. 2attery failure can
result in lost sensor node. )his leads to a gradual degradation of the net!or" as individual nodes
are lost. *et!or" #aths brea" and ga#s a##ear in the coverage area. A S* needs to ada#t to the
changes, recover from losses and be sel'#protected. )his can be achieved by renegotiating
net!or" routes, monitoring voltage levels !ithin sensor node, controlling each node by an agent
or base station and u#on failure activating redundant nodes to re#lace damaged ones, or by
informing some higher-level entity !hich can #rovide assistance.
As discussed in re-uirements, ma$imum efficiency needs to be gained from the available
energy as the available energy at each sensor node is limited. Sensing, .rocessing and data
transfer #hases re-uire lot of energy so each node should be able to sense #rocess and transfer
data intelligently hence sel'#optimi)ation is an im#ortant trait for S* #rotocols. Energy
savings can be achieved by #utting the nodes into a lo! #o!er slee# mode, ready to be
reactivated !hen the need arises. 0or e$am#le, sensors may decrease their duty cycles !hen
there is no significant change in their sensor readings. )his results in less #o!er consum#tion in
the sensors. Also, !hen neighboring sensors re#ort environmental changes, a sensor may dra!
inference from the re#orts and increase its duty cycle to be more !atchful for a #otential local
environmental change in the future. Bo!ever, there e$ists a trade-off in that the com#utational
cost of a globally-o#timal solution such as this is often com#utationally intractable, !hether by
C-bit nodes or ?=-bit base-stations.
All basic S* self-management #rinci#les com#ly !ith the conce#t of autonomic com#uting.
So %23 autonomic com#uting #rinci#les can be a##lied to !ireless sensor net!or"s to get the
desired functionality in vastly gro!ing sensor net!or" a##lications.
*. Autonomic Wireless Sensor Network !anagement Arc"itectures #
As discussed in section 9.8, the basic Autonomic Com#uting model only #rovides the conce#tual
guidance on designing self-managed systems and needs to be ma##ed to an im#lementable
management and control architecture for Autonomic *et!or"s. An architecture for Autonomic
communication and net!or"ing is an area of research lately and many architectures are #ro#osed
and being develo#ed. All these architectures aim to #roduce an architectural design that enables
fle$ible, dynamic and fully autonomic formation of large-scale net!or"s in !hich the
functionalities of each constituent net!or" node are also com#osed in an autonomic fashion.
3oreover, these architectures also su##ort mobile nodes and multi#le administrative domains so
these can be a##lied to !ireless sensor net!or"s for achieving desired goals and meet above
mention challenges.
0ollo!ing is the brief discussion of some visions for the design of an efficient management
architecture for S*s based on to# of the basic autonomic com#uting architecture.
*.1. Ser+ice#,riented Arc"itecture
Service-oriented architecture 'S1A +ef 4967( is an a##roach to build distributed systems that
deliver a##lication functionality as services to end-user a##lications or to build other services. %t
decom#oses the design of large com#le$ a##lication, and middle!are architecture into various
reusable services or function units. %n S1A the service re-uester has no "no!ledge of the
technical details of the #rovider:s im#lementation, such as the #rogramming language,
de#loyment #latform, and so forth. )he service re-uester ty#ically invo"es o#erations by !ay of
messages -- a re-uest message and the res#onse -- rather than through the use of A.%s or file
formats. )hus, the a##lication develo#ers only need to concern the o#erational descri#tion of the
service !hich allo!s soft!are on each side of the conversation to change !ithout im#acting the
other.
So far, the im#lementation and design of S1A is mostly de#endent on eb Services !ith
standardi&ed !eb technologies such as S,L, 1DSA. As a result, it is not directly a##licable to
all of those com#le$ technologies on those resource-constrained sensor nodes. 3A**A 46E7 has
#resented some initial ideas of using the conce#t of service semantics from S1A.
0igure =. 2asic 3odel of 3A**A Architecture '+ef 46E7(
%n 3A**A, all the management function units sit at the lo!est level of management
architecture. )hey are designed !ith s#ecific im#lementation for individual ob/ectives in
consideration of uni-ue features of S*. A service, at the to# layer, can use one or more of
those management functions. ,ifferent services can share the same functions, but still concern
each individual given as#ect based on the #olices and net!or" state obtained from S* models.
)he basic model of 3A**A architecture is as sho!n in figure = '+ef 46E7(
0urthermore, S1A can s#ecially deal !ith S* uni-ue as#ects such heterogeneity, mobility and
ada#tation, and offers seamless management integration in the !ireless environments. Although
the s#ecial features of S1A are marvellous, there is still a large amount of research challenge
needed to address before the conce#ts of S1A can be a##ro#riately a##lied into S*s.
*.2. -olicy Based Arc"itecture
.olicy-based management has #resented its robust ability to su##ort designing of self-ada#tive
decentrali&ed management service in S*s. ,avy S. et al. 4957 #ro#osed an autonomic
communications architecture that manages com#le$ity through #olicy-based management by
incor#orating a shared information model integrated !ith "no!ledge-based reasoning
mechanisms to #rovide self-governing behavior.
)he architecture is organi&ed using four distinct architectural constructs i.e. Shared $nformation
7irtual Software $nfrastructure and .olicy as sho!n in figure >.

0igure >. .ro#osed .olicy 2ased Autonomic Architecture '+ef 4957(
)he shared information over the net!or" is managed through a virtual soft!are !hich su##ort
autonomic functionality for different heterogeneous net!or"s and com#onents combined !ith
net!or" infrastructure !hich include net!or" elements and other com#uting devices. All these
three modules are governed by #olicy module. 4957
)his model is based on three im#ortant conce#ts of autonomic com#uting: '6( the sharing and
reusing of common information and "no!ledge, '9( the a##lication of machine learning and
"no!ledge-based reasoning to guide the changes in behavior of the system, and '8( an e$tensible
and fle$ible governance model that forms a closed control loo# that learns from its decisions.
Similarly, in 3A**A 46E7, #olicies describe a set of desired behaviours of management
com#onents 'e.g. manager and agent( for indicating the real-time o#erations. 2ased on #olices,
managers and agents can interact !ith each other in a coo#erative fashion to achieve a desired
overall management goal such as form grou#s of nodes, control net!or" density, and "ee# the
coverage of the S* area.
. $outing -rotocols 'or Autonomic Wireless Sensor Networks
*o! let:s analy&e fe! !ell-"no!n routing #rotocols for !ireless sensor net!or"s and their
suitability, #ros and cons for Autonomic S*:s.
..1 /looding
0looding 49=7 is an old routing mechanism used in !ireless net!or"s that may also be used in
Autonomic !ireless sensor net!or"s. %n flooding, a node sends out the received data or the
management #ac"ets to its neighbors by broadcasting or flooding, unless a ma$imum number of
ho#s for that #ac"et are reached or the destination of the #ac"ets is arrived. )his method
guarantees the delivery of the #ac"et to the destination. Bo!ever< there are some deficiencies
and disadvantages of flooding techni-ue 49=7:
- %m#losion of ,ata #ac"ets: 0looding may cause the %m#losion effect for data #ac"ets and also
for ACF #ac"ets if used any. 1ne #ac"et ta"es multi#le routes and multi#le hosts can deliver the
same #ac"et to destination. ,estination has to im#lement a se#arate mechanism for du#licate
su##ression also lot of band!idth and resources are !asted in transmitting same #ac"et through
multi#le hosts so this techni-ue may not be suitable for large ireless sensor net!or"s.
- 1verla#: if t!o sensor nodes cover an overla##ing measuring region, both of them !ill
senseGdetect the same data. As a result, their neighbor nodes !ill receive du#licated data or
messages. 1verla##ing is a function of both the net!or" to#ology and the ma##ing of sensed
data to sensor nodes.
- +esource utili&ation: %n flooding, nodes do not ta"e into account the amount of energy resource
available to them at a given time. An Autonomic S* #rotocol must be energy resource-a!are
and ada#ts its sensing, communication and com#utation to the state of its energy.
..2 0ossiping
Dossi#ing #rotocol is an alternative to flooding mechanism. %n Dossi#ing 49>7, nodes for!ard
incoming #ac"ets to a randomly selected neighbor node. 1nce a gossi#ing node receives the
messages, it for!ards the data bac" to that neighbor or to another one randomly selected
neighbor node and in this !ay route from source to destination is created. )his techni-ue assists
in energy conservation by randomi&ation.
Although, gossi#ing can solve the im#losion #roblem, it can not avoid the overla##ing #roblem.
1n the other hand< gossi#ing distribute information slo!ly, this means it consumes energy at a
slo! rate, but the cost is long-time #ro#agation is needed to send messages to all sensor nodes so
it may not be the best suitable techni-ue for Autonomic ireless Sensor net!or"s.
.. S-IN
Fuli" et al. #ro#osed a family of ada#tive #rotocols for S*s, called S.%* 'Sensor .rotocols for
%nformation via *egotiation( 49?7. )heir design goal is to avoid the dra!bac"s of flooding
#rotocols by utili&ing data negotiation and resource-ada#tive algorithms. *odes running a S.%*
communication #rotocol name their data using high-level data descri#tors, called meta-data.
)hey use meta-data negotiations to eliminate the transmission of redundant data throughout the
net!or". %n addition, S.%* nodes can base their communication decisions both u#on a##lication-
s#ecific "no!ledge of the data and u#on "no!ledge of the resources that are available to them.
)his efficient distribution of data by sensors !ith limited energy su##ly com#lies !ith the
Autonomic Sensor net!or" re-uirements and can be very effective under small net!or"s.
S.%* is designed based on t!o basic ideas< '6( to o#erate efficiently and to conserve energy by
sending metadata 'i.e., sending data about sensor data instead of sending the !hole data that
sensor nodes already have or need to obtain(, and '9( nodes in a net!or" must be a!are of
changes in their o!n energy resources and ada#t to these changes to e$tend the o#erating lifetime
of the system. S.%* has three ty#es of messages, namely, A,V, +EH, and ,A)A.
- A,V: !hen a node has data to send, it advertises via broadcasting this message containing
meta-data 'i.e., descri#tor( to all nodes in the net!or".
- +EH: an interested node sends this message !hen it !ishes to receive some data.
- ,A)A: ,ata message contains the actual sensor data along !ith meta-data header.
S.%* is a data-centric routing #rotocol !here the sensor nodes send A,V message via
broadcasting for the data they have and !ait for +EH messages from interested sin"s or nodes.
S.%* has some advantages in solving the #roblems associated !ith classic flooding #rotocols,
and ada#tive to to#ological changes, it has its o!n dra!bac"s li"e< '6( S.%* is not scalable, '9( if
the sin" is interested in too many events, this could ma"e the sensor nodes around it de#lete their
energy, and '8( S.%*Is data advertisement techni-ue can not guarantee the delivery of data if the
interested nodes are far a!ay from the source node and the nodes in bet!een are not interested in
that data.
..( &irected &i''usion
,irected diffusion 49@7 is most effective data dissemination and aggregation #rotocol. %t is a data-
centric and a##lication a!are routing #rotocol for ireless Sensor *et!or"s. %t aims at naming
all data generated by sensor nodes by attribute-value #airs. ,irected diffusion consists of several
elements< first of all, naming< !here tas" descri#tors, sent out by the sin" or ,ata receiver, are
named by assigning attribute-value #airs. Secondly, interests and gradients< the named tas"
descri#tion constitutes an interest that contains timestam# field and several gradient fields.
Each leaf node and intermediate nodes store the interest in their interest cache. As the interests
#ro#agate throughout the net!or", the gradients from the source bac" to the sin" are set u#.
)hirdly, data #ro#agation, !hen the source has data for the interest, it sends out the data to the
interest 'i.e., sin"( along the interestIs gradient #ath !hich can be chosen as the shortest ho# #ath
or shortest time #ath derived from the re-uest #ac"et. 0ourthly, after the interest 'sin"( starts
receiving lo! rate data events, it reinforce one #articular neighbor to dra! do!n higher -uality
'higher data rate( events. )his feature of directed diffusion is achieved by data-driven local rules.
,irected diffusion assists in saving sensorsI energy by selecting good #aths by caching and
#rocessing data in-net!or" since each node has the ability for #erforming data aggregation and
caching. 1n the other hand< ,irected diffusion has its limitations such as< im#lementing data
aggregation re-uires de#loyment of synchroni&ation techni-ues !hich is not reali&able in S*s.
Also, the overhead in data aggregation involves recording information.
)hese t!o dra!bac"s may contribute to the cost of sensor node hence cost is the tradeoff for
#erformance in ,irected diffusion techni-ue, !hich may be acce#table for some autonomic
!ireless sensor net!or"s.
..* 12AC3
LEACB 'Lo! Energy Ada#tive Clustering Bierarchy( 49C7 is a self-organi&ing, ada#tive
clustering-based #rotocol that uses randomi&ed rotation of cluster-heads to evenly distribute the
energy load among the sensor nodes in the net!or". LEACB based on t!o basic assum#tions:
'a( base station is fi$ed and located far a!ay from the sensors, and 'b( all nodes in the net!or"
are homogeneous and energy-constrained. )he idea behind LEACB is to form clusters of the
sensor nodes de#ending on the received signal strength and use local cluster heads as routers to
route data to the base station. )he "ey features of LEACB are:
- Locali&ed coordination and control for cluster set-u# and o#eration.
- +andomi&ed rotation of the cluster Jbase stationsJ or Jcluster headsJ and the corres#onding
clusters.
- Local com#ression to reduce global communication.
%n LEACB, the o#eration is se#arated into fi$ed-length rounds, !here each round starts !ith a
setu# #hase follo!ed by a steady-state #hase. )he duration of a round is determined #riori.
Although, LEACB has sho!n good features to sensor net!or"s, it suffers from the follo!ing
dra!bac"s:
- %t can not be a##lied to time-constrained a##lication as it results in a long latency.
- )he nodes on the route a hot s#ot to the sin" could drain their #o!er fast. )his #roblem is
"no!n as Jhot s#otJ #roblem.
- )he number of clusters may not be fi$ed every round.
- %t can not be a##lied to large sensor net!or"s.
)herefore, for Autonomic !ireless sensor net!or"s !ith stable and fi$ed homogeneous nodes
the LEACB #rotocol !ill give good #erformance. 0or a Autonomic Sensor *et!or" !ith
stationary, battery #o!ered nodes it !ould be effective to use clustered based #rotocol li"e
LEACB, the most obvious reason is, its advantages such as reduced control messages, band!idth
reusability, enhanced resource allocation, im#roved #o!er control and lest !astage of energy.
... -20ASIS
.EDAS%S '.o!er-Efficient DAthering in Sensor %nformation Systems( is a greedy chain-based
#o!er efficient algorithm 49E7. .EDAS%S is based on t!o ideas i.e. Chaining, and ,ata 0usion. %t
uses same techni-ue as LEACB 'the scenario and the radio model in .EDAS%S are the same as
in LEACB(.
%n .EDAS%S, each node can ta"e turn of being a leader of the chain, !here the chain can be
constructed using greedy algorithms that are de#loyed by the sensor nodes. .EDAS%S assumes
that sensor nodes have a global "no!ledge of the net!or", nodes are stationary 'no movement of
sensor nodes(, and nodes have location information about all other nodes. .EDAS%S #erforms
data fusion e$ce#t the end nodes in the chain. .EDAS%S out#erforms LEACB by eliminating the
overhead of dynamic cluster formation, minimi&ing the sum of distances that non leader-nodes
must transmit, limiting the number of transmissions and receives among all nodes, and using
only one transmission to the 2ase Station #er round.
As it is similar as LEACB #rotocol, .EDAS%S also suffers from same #roblems as LEACB.
Additionally, .EDAS%S does not scale, so can not be a##lied to sensor net!or" !here global
"no!ledge of the net!or" is not easy to get.
..4 02A$
DEA+ 'Deogra#hical and Energy A!are +outing( 4857 is a recursive data dissemination #rotocol
for S*s. %t uses energy a!are and geogra#hically informed neighbor selection heuristics to
route a #ac"et to the targeted region. ithin that region, it uses a recursive a geogra#hic
informed mechanism to disseminate the #ac"et. DEA+, li"e other sensor net!or"s #rotocols,
develo#ed according to some assum#tions in mind:
- Sensor nodes are static 'i.e., immobile(.
- )here is an e$istence of a locali&ation system that enables each node to "no! its current
#osition.
- Sensor nodes are energy-constrained accom#anied !ith location information about all other
nodes 'i.e., each node "no!s its location and its energy level, and its neighborIs location and
remaining energy level.
- )he lin" that connects nodes is bi-directional.
DEA+ has t!o #hases: '6( for!arding the #ac"ets to!ard the targeted region, and '9( for!arding
the #ac"ets !ithin the targeted region +.
Although DEA+ reduces the energy consum#tion for the route set u#. %t is not scalable and does
not su##ort data diffusion.
2ased on the analysis, com#atibility survey of the e$isting #rotocols for ireless Sensor
*et!or"s, !e can conclude that some of the #rotocols can more or less be a##lied for routing in
Autonomic ireless Sensor *et!or"s !ith fe! modifications de#ending u#on the net!or"
structure and functionality. 1verall, there are some "ey features, an efficient routing #rotocol for
Autonomic ireless Sensor *et!or"s should have are: 4867
- &ata Aggregation5 +educing the data si&e -uic"ly using com#utation !ill #lay a "ey role in
su##orting efficient -uery #rocessing, and reducing the overall net!or" overhead. Bence saving
the #o!er.
- &ynamic clustering5 ,ynamic clustering architecture is very im#ortant because such
architecture !ill #reclude cluster heads from de#leting their energy -uic"ly. Bence, long
net!or"Is lifetime.
- 6"res"old 'or sensor nodes on data transmission and dissemination5 this !ill hel# in saving
energy by reducing unnecessary transmissions 'i.e., redundancy( and giving the net!or" long
lifetime.
- $andomi)ed pat" selection5 multi-#ath selection to destination could im#rove fault tolerance
and handle the overhead of net!or" load.
- !obility5 most of the current #rotocols assume that sensor nodes are static 'i.e., immobile(.
Bo!ever, for some a##lications, nodes need to be mobile. Bence, ne! routing algorithms are
needed to handle the mobility and net!or" to#ology changes.
- Sel'#con'iguration5 since sensor nodes are #rone to failure due to some factors or ne! sensor
nodes may /oin the net!or", an u#date, self-configuration, self-healing, and ada#tation to
changes in net!or" to#ology or environmental changes should be considered.
- Security5 there is a des#erate need to develo# distributed security a##roaches for !ireless
sensor net!or". Bence, achieving secure routing.
- Huality-of-Service, de#endability, and locali&ation need to be considered and given more
attention.
- )ime synchroni&ation.
4. Brie' ,+er+iew o' researc" pro7ects on Autonomic Networks
Bere is a brief overvie! of the current research #ro/ects based on Architecture for Autonomic
*et!or" communication and Self-3anagement !hich !ill serve as guidelines for Autonomic
S*:s and !ill bring revolution to S*:s and its a##lications.
4.1. Bison
2%S1* !as a three-year #ro/ect funded by the Euro#ean Commission. 2%S1* aimed
confronting the com#le$ity e$#losion #roblem by building robust *et!or" %nformation Systems
that are self-organi&ing and self-re#airing.
2%S1* develo#ed techni-ues and tools for building robust, self-organi&ing and ada#tive
*et!or" %nformation System as ensembles of autonomous agents by dra!ing ins#iration from
biological #rocesses and mechanisms li"e ant colonies for routing in overlay net!or"s using
s!arm intelligence, lifecycle of ,ictyostelium for load balancing, e#idemics for aggregation and
immune system for search.
2%S1* e$#lored the use of ideas derived from com#le$ ada#tive systems 'CAS( to enable the
construction of robust and self-organi&ing information systems for de#loyment in highly
dynamic net!or" environments. )he #ro/ect #ro#osed solutions to im#ortant #roblems arising in
overlay net!or"s and mobile ad-hoc net!or"s by develo#ing algorithms for routing in mobile
ad-hoc net!or"s, to#ology control in sensor net!or"s along !ith data aggregation and content
search algorithms for #eer to #eer net!or"s. 4667
4.2. ANA 8Autonomic Network Arc"itecture9
A*A frame!or" is built on the ob/ective to #rovide an architectural frame!or" that allo!s the
accommodation of and communication bet!een various net!or"s, ranging from small scale
.ersonal Area *et!or"s, through '3obile( Ad hoc *et!or"s and s#ecial #ur#ose net!or"s such
as Sensor *et!or"s, to global scale net!or"s, in #articular the %nternet. A*A frame!or"
s#ecifies ho! net!or"s interact.
A*A introduces the core conce#t of Jnet!or" com#artments.J )he com#artment abstraction
allo!s atomi&ation or decom#osition of communication systems and net!or"s into smaller and
more easily manageable units. 0or e$am#le, com#artments !ill allo! decom#osition of today:s
global %. net!or" into a##ro#riate sub-net!or"s, !hich can be managed more autonomously
from the overall net!or" 'e.g., a different addressing or routing scheme can be a##lied inside
each com#artment(. 46=7
0igure ?. A*A 0rame!or" and *et!or" com#artments '+ef 46=7(
A 'net!or"( com#artment im#lements the o#erational rules and administrative #olicies for a
given communication conte$t. Com#artments ty#ically #erform functions li"e registration and
degradation, #olicy enforcement, identifier management and resolution and +outing. 46=7
Addressing and naming are left to com#artments. )he main advantages of this a##roach are:
*o need to im#ose a uni-ue !ay to resolve names and manage a uni-ue global addressing
scheme. %t is o#en to future addressing and naming schemes.
4.. 3aggle
Baggle is a ne! autonomic net!or"ing architecture designed to enable communication in the
#resence of intermittent net!or" connectivity, !hich e$#loits autonomic o##ortunistic
communications 'i.e., in the absence of end-to-end communication infrastructures(. Baggle node
architecture ta"es ins#iration from human communication model. 46>7
)he main com#onents of Baggle are:
A re+olutionary paradigm 'or autonomic communication, based on advanced local
for!arding and sensitive to realistic human mobility
A simple and power'ul arc"itecture oriented to o##ortunistic message relaying, and
based on #rivacy, authentication, trust and advanced data handling
An o#en environment for the easy #roliferation of a##lications and services.
4.(. CASCA&AS
CASCA,AS 'Com#onent-!are for Autonomic Situation-a!are Communications, and
,ynamically Ada#table Services( is an ongoing #ro/ect li"e A*A and Baggle.
)he overall goal of CASCA,AS is identifying, develo#ing, and evaluating architectures and
solutions based on a general-#ur#ose com#onent model for autonomic communication services<
s#ecifically in such conte$t autonomic service com#onents autonomously achieve self-
organi&ation and self-ada#tation to!ards the #rovision of ada#tive and situated communication-
intensive services.
CASAC,AS a##roach is based on four "ey scientific #rinci#les i.e. situation a!areness,
semantic self organi&ation, self similarity and Autonomic com#onent a!areness around !hich
the future communication services infrastructures should be designed and built. 46?7
:. Applications ; 'uture work
)he a##lications for S*s are many and varied. )hey are used in commercial and industrial
a##lications to monitor data that !ould be difficult or e$#ensive to monitor using !ired sensors.
)y#ical a##lications of S*s include monitoring, trac"ing, and controlling. Some of the s#ecific
a##lications are habitat monitoring, ob/ect trac"ing, nuclear reactor controlling, fire detection,
traffic monitoring and so on.
6. ireless sensor net!or"s are currently being used for intrusion detection by forming a
#erimeter around a secure area and monitoring the #rogression of intruders '#assing information
from one node to the ne$t(. S*:s could be further de#loyed in 3ilitary a##lications such as
hostile trac"ing and surveillance, s#y monitoring.
9. 1ther ma/or current a##lication of S* include environment monitoring and a##lications
such as animal trac"ing, flood detection and !eather #rediction and forecasting and commercial
a##lications li"e seismic activities monitoring and #rediction. 3any !eather forecasting
!ebsites use S* technology for retrieving !eather details in remote inhibited areas. 4987
8. Significant amount of the technology and a##lications are already in e$istence for monitoring
activities in home along !ith intrusion detection by e-ui##ing a home !ith a suitable sensor-
laden infrastructure.
=. S*:s are used !idely in automation and control and Artificial intelligence a##lications li"e
+obotics.
> Sensor net!or"s are increasingly being used in Bealth a##lications for monitoring changes in
#atient:s health, behaviour and heart rate.
2y continuously monitoring the #rogressive disease, o##ortunities for actively intervening to aid
the #atient may be identified. )he Ambient Assisted living technologies are in e$istence, !hich
use S* elements to assist the #atient.
+esent research #ro/ect at ayne State University and the Fresge Eye %nstitute develo#ed
artificial retina using ireless 2iomedical sensors. )he #ro/ect aimed to build a chronically
im#lanted artificial retina !ith sufficient visual functionality to allo! #ersons !ithout vision or
!ith limited vision to KseeL at an acce#table level. 46@7
3oreover, this ireless biomedical sensor technology can be effectively used to treat diabetes,
by #roviding a more consistent, accurate, and less invasive method for monitoring glucose levels.
Currently, to monitor blood glucose levels, a lancet is used to #ric" a finger< a dro# of blood is
#laced on a test stri#, !hich is analy&ed either manually or electronically. )his constant #ric"ing
several times a day over a #eriod of years can damage the tissue and blood vessels in that area.
As #ro#osed by Sch!iebert et al. 46@7, ireless biomedical sensors could be im#lanted in the
#atient once. )he sensor !ould monitor the glucose levels and transmit the results to a
!rist!atch dis#lay.
ireless biomedical sensors may #lay a "ey role in early detection of Cancer. As discussed in
46@7, cancer cells e$ude nitric o$ide, !hich affects the blood flo! in the area surrounding a
tumor. A sensor !ith the ability to detect these changes in the blood flo! can be #laced in
sus#ect locations. %t is li"ely that any abnormalities could be detected much sooner !ith the
sensors than !ithout.
+0%,, video and various "inds of embedded sensors can be used to trac" and monitor the
#atient in their everyday activities. )his information can be #rocessed and relayed to medical
#ersonnel. .atientIs routine can be assembled over the #eriod of time and deviations from this
may be recogni&ed and analy&ed.
<. Conclusion
ireless Sensor *et!or" technology offers significant #otential in numerous a##lication
domains. Diven the diverse nature of these domains, it is essential that S*s #erform in a
reliable and robust fashion. % believe, !ireless sensor net!or" has #roved its usage in the future
distributed com#uting environment. Bo!ever, there are significant amount of technical
challenges and design issues those needs to be addressed. 1ne of the biggest challenges is the
designing of efficient net!or" management architecture to continuously su##ort S*s for
#roviding services for various sensor a##lications. )he uni-ue features of S*s ma"e the design
and im#lementation of such management architecture different enough from the traditional
net!or"s !hich can be satisfied by conce#t of Autonomic Com#uting. )here is still no #articular
generic net!or" management architecture so ta"ing ins#iration from %23s Autonomic
Com#uting conce#t and 2iological neural net!or" system many different research #ro/ects are
currently being e$ecuted.
%n this #a#er, !e discussed conce#ts of Autonomic com#uting, ireless Sensor *et!or"s
'S*:s(. ,esign criteria for S* and ho! it matches basic Autonomic #rinci#les. )hen !e
overvie!ed fe! architectures and routing #rotocols suitable for S* and ongoing research !or"
of Autonomic communication and net!or" management architectures !hich can be a##lied to
S*s. 0inally, !e summari&ed some of the S* a##lications along !ith future usages.
$e'erences
4567 .. Born, 8!utonomic )omputing+ $%&s .erspective on the State of $nformation
(echnology9, 1ct. 9556. Available from the orld ide eb '(:
htt#:GG!!!.research.ibm.comGautonomicG
4597 %23 and autonomic com#uting, 8!n architectural blueprint for autonomic computing9
A#ril 9558. Available from the : htt#:GG!!!-
58.ibm.comGautonomicG#dfsGAC!#0inal.#df
4587 Fe#hart M, and Chess ,, K(he 7ision of !utonomic )omputingL Com#uter 3aga&ine, %EEE,
9558. Available from :
htt#:GG!!!.research.ibm.comGautonomicGresearchG#a#ersGACNVisionNCom#uterNManN9558.#
df
45=7 %. A"yildi&, . Su, O. San"arasubramaniam, and E. Cayirci, L! survey on Sensor Networks,L
%EEE Communications 3aga&ine, vol. =5, %ssue: C, ##. 659-66=, August 9559. Available
from : htt#:GGciteseer.ist.#su.eduGa"yildi&59survey.html
45>7 Ou 3eng/ie, 3o"htar B., 3erabti 3., K! Survey of Network &anagement !rchitecture in
Wireless Sensor NetworkL Available from :
!!!.cms.liv/m.ac.u"GsenmanG.a#ersG955?-5E8.#df
45?7 C. %ntanagon!i!at, +. Dovindan, ,. Estrin, M. Beidemann, and 0. Silva, L4irected 4iffusion
for Wireless Sensor Networking,L %EEEGAC3 )ransactions on *et!or"ing, vol. 66, ##. 9-6?,
0eb. 9558. Available from : htt#:GGciteseer.ist.#su.eduGintanagon!i!at55directed.html
45@7 1IBare D.3..., 1IDrady 3.M., ,. 3arsh, +u&&elli A. D. and )ynan +. K!utonomic Wireless
Sensor Networks+ $ntelligent. :bi"uitous SensingL Available from :
!!!.cs.ucd.ieGcs#rismG#ublicationsG#ub955?GA*%.LA5?.#df
45C7 ,. 3arsh, +. )ynan, ,. 1:Fane and D. 3. .. 1:Bare, KAutonomic ireless Sensor
*et!or"s,L 3ngineering !pplications of !rtificial $ntelligence, vol 6@-@, ##. @=6-@=C,
1ctober 955=. Available from : htt#:GGtinyoside.ucd.ieG#ublicationsGretrieve.#h#P
#ublicationQtynanAutonomic5>.#df
45E7 Ou Cheng et al., K! generic architecture for autonomic service and network managementL,
Com#uter Communications '955?(, doi:65.656?G/.comcom.955?.5?.56@. Available from
: htt#:GG!!!.ece.iit.eduGRyuchengGOChengNCom#Com.#df
4657 .. 2oonma, .. Cham#rasert and M. Su&u"i, J%iSN3(+ ! %iologically-$nspired !rchitecture
for Wireless Sensor Networks,J %n .roc. of the 1nd $333;$!0$! $nternational )onference on
!utonomic and !utonomous Systems ($333;$!0$! $)!S), Santa Clara, CA, Muly 955?.
Available from : htt#:GG!!!.cs.umb.eduGR/$sG#ubGicas-bisnet.#df
4667 2%S1*: 2iology-%ns#ired techni-ues for Self-1rgani&ation in dynamic *et!or"s .ro/ect.
"ttp5==www.cs.unibo.it=bison=inde>.s"tml
4697 Shen C, .esch ,, %rvine M.,L! #ramework for Self-&anagement of ,ybrid Wireless
Networks :sing !utonomic )omputing .rinciplesL %n #roceedings of the 8rd Annual
Communication *et!or"s and Services +esearch Conference 'C*S+I5>( - Volume 55.
Available from : htt#:GG#ortal.acm.orgGcitation.cfmP
idQ65?C>58.65?C@66ScollQDU%,ESdlQDU%,ESC0%,QE=?@@==SC0)1FE*Q==C89C68
4687 Colan, 3. Service-1riented Architecture e$#ands the vision of eb services, .art 6. Mune,
955=
46=7 A*A: Autonomic *et!or" Architecture .ro/ect. htt#:GG!!!.ana-#ro/ect.orgG
46>7 Baggle .ro/ect. htt#:GG!!!.haggle#ro/ect.org
46?7 CASCA,AS #ro/ect. htt#:GG!!!.cascadas-#ro/ect.org
46@7 Sch!iebert L., Du#ta S., einmann M., K0esearch )hallenges in Wireless Networks of
%iomedical SensorsL in #roceedings of the @th annual international conference on 3obile
com#uting and net!or"ing. Available from : htt#:GG#ortal.acm.orgGcitation.cfmP
idQ8C6?E9
46C7 Autonomic Com#uting i"i. htt#:GGen.!i"i#edia.orgG!i"iGAutonomicNcom#uting
46E7 Linnyer 2eatrys +ui&, M.3.S.*., Antonio A.0. Loureiro, 8&!NN!+ ! &anagement
!rchitecture for Wireless Sensor Networks.9 %EEE Communications 3aga&ine, 9558. =6'9(:
#. 66?-69>. Available from : htt#:GG!!!.lisha.ufsc.brGRlucasGdocsG+ui&-9558.#df
4957 ,avy S. et al. K.olicy-%ased !rchitecture to 3nable !utonomic )ommunications9. Available
from :
htt#:GGtech#ubs.motorola.comGdo!nloadG%.C135556=6=55,G%.C135556=6=55,.#df
4967 Colan, 3. Service-1riented Architecture e$#ands the vision of eb services, .art 6. Mune,
955= Available from : htt#:GG!!!-69C.ibm.comGdevelo#er!or"sGlibraryG!s-
soaintro.html
4997 Dianni A., ,i Caro, 0rederic" ,ucatelle, Luca 3. Dambardella, K2%S1*: 2iology-%ns#ired
techni-ues for Self-1rgani&ation in dynamic *et!or"sL Available from :
htt#:GG!!!.idsia.chGRfrederic"Gbison.#df
4987 ireless Sensor *et!or"s i"i. htt#:GGen.!i"i#edia.orgG!i"iGirelessNsensorNnet!or"
49=7 . +. Bein&elman, M. Fuli", and B. 2ala"rishnan, J!daptive .rotocols for $nformation
4issemination in Wireless Sensor NetworksJ in #roc. AC3 3obiCom IEE, Seattle, A,
6EEE. Available from :
htt#:GG!!!.cs.hu/i.ac.ilGlabsGdanssGsensorGsensorsG"uli"NEEada#tive#rotocols.#df
49>7 S. 3. Bedetniemi, S. B. Bedetniemi, and A. Liestman, J! Survey of <ossiping and
%roadcasting in )ommunication Networks,J *et!or"s, vol. 6C, 6ECC.
49?7 M. Fuli", . +. Bein&elman, and B. 2ala"rishnan, JNegotiation base protocols for
4isseminating $nformation in Wireless Sensor Networks,J ireless *et!or"s, vol. C, ##.
6?E-6C>, 9559. Available from : htt#:GGciteseer.ist.#su.eduG"uli"EEnegotiationbased.html
49@7 C. %ntanagon!i!at, +. Dovindan, ,. Estrin, M. Beidemann, and 0. Silva, J4irected 4iffusion
for Wireless Sensor Networking,J %EEEGAC3 )ransactions on *et!or"ing, vol. 66, ##. 9-6?,
0eb. 9558. Available from : htt#:GG!!!.isi.eduGR/ohnhG.A.E+SG%ntanagon!i!at58a.#df
49C7 . Bein&elman, A. Chandra"asan, and B. 2ala"rishnan, J3nergy-3fficient )ommunication
.rotocol for Wireless &icro-Sensor Networks,J in .roc. of the 88rd Annual Ba!aii %nternational
Conf. on System Sciences, ##. 855>- 856=, 9555. Available from :
htt#:GG#dos.csail.mit.eduGdecoutoG#a#ersGhein&elman55.#df
49E7 S. Lindsey, C. S. +aghavendra, J.3<!S$S+ .ower-3fficient <athering in Sensor
$nformation Systems,J #resented at .roc. of %EEE Aeros#ace Conference, 3ontana, 9559.
Available from : htt#:GGceng.usc.eduGRraghuG#egasisrev.#df
4857 O. Ou, +. Dovindan, ,. Estrin, J<eographical and 3nergy !ware 0outing+ ! 0ecursive 4ata
4issemination .rotocol for Wireless Sensor Networks,J UCLA Com#uter Science
,e#artment UCLA-CS, )+-56-5598, 3ay, 9556. Available from :
htt#:GG!!!.#arc.comG&haoGstanford-cs=9CGreadingsG*et!or"ingGEstrinNgeo-routing56.#df
4867 Al-1baisat O, 2raun + K1n ireless Sensor *et!or"s: Architectures, .rotocols,
A##lications, and 3anagementL Aus!ireless Conference 955?. Available from :
htt#:GGe#ress.lib.uts.edu.auGds#aceGbitstreamG9655G6?5G9G>ENAl-1baisat.#df