_\

A Simple and Efficient Method to Mitigate the Hot SpotProblem in Wireless Sensor Networks

Helena Rivas, Thiemo Voigt, Adam Dunkels

Swedish Institute ofComputer Science, Box 1263, SE-164 29 Kista, Sweden

{hefena, thiemo, adam}Gsics. se

Abstract. Much work on wireless sensor networks deals with or considers the

hot spot problem, i.e., the problem that the sensor nodes closest to the base station

are critical for the lifetime of the sensor network because these nodes need to

relay more packet than nodes further away from the base station. Since it is often

assumed that sensor nodes will become inexpensive, a simple solution to the hot

spot problem is to place additional sensor nodes around the base stations. Using a

simple mathematical model we discuss the possible performance gains of adding

these supplementary nodes. Our results show that for certain networks only a

limited number of additional nodes are required to fourfold network lifetime. We

also show that the possible gain depends heavily on the fraction of nodes alreadypresent in the vicinity of the base station.

I Introduction

The main task of most wireless sensor networks is to collect data and send it in a multi-hop fashion to a base station. While forwarding data in a multi-hop fashion to the base

station is often more energy-efficient than transmitting the data directly from the sens-

ing node to the base station, a potential disadvantage of the multi-hop strategy is thatthe nodes close to the base station must forward much more packets than nodes furtheraway from the base station. Therefore, these nodes "typically die at an early stage" [7].This is sometimes called the hot spot problem []. Without adding extra nodes or redis-tributing the available energy, this problem is hard to solve. For example, Perillo et al.

have shown that varying the transmission power of nodes, even considering unlimitedtransmission ranges, does not solve the hot spot problem [7].

At the same time, it is also envisioned that sensor nodes will become "extremelyinexpensive" [5]. While beyond a certain node density, adding additional nodes does notprovide any improvement regarding sensing, communication or coverage [3], addingnodes might obviously help to increase the lifetime of a sensor network while providingthe same service to its users, i.e. leveraging sensor values from the same number ofnodes.

In this paper, we study the benefit of adding extra nodes to a sensor network usinga mathematical model we developed previously [2]. Our results show that for certainnetworks only a limited number of additional nodes are required to fourfold networklifetime. We also show that the possible gain depends heavily on the fraction of nodesalready present in the vicinity of the base station.

NE RK ARC HITECTURESD PROTOCOL STACK

Jun Zheng

Southeast U niv ersity, China

2.1 INTROD

Network architectures d protocols are important aspects in the design of wire-less sensor networks ( ) [1]. Due to the severe energy constraint of sensornodes, network archi design has a big impact on the energy consumptionand thus the operati lifetime of the whole network. On the other hand, a

of a large number of sensor nodes that are denselvion and collaborate to accomplish a sensing task. Itprotocols to implement various network control andexample, synchronization, self-confi.guration, medium

r networks because they do not consider the energy,constraints in sensor nodes. On the other hand, most

Perspective, Edited by Jun Zheng and Abbas Janralipourand Electronics Engineers

sensor network consisdeployed in a sensingrequires a suite ofmanagement lunctiaccess control, routing, ta aggregation, node localization, and network securitv.However, existing net protqiols for traditional wireless networks. forexample, cellular s and mobile ad hoc networks (MANETs), cannot beapplied directly to sencomputation, andsensor networks are ication specific and have different application require-ments. For these new suite of network protocols is required, which takeinto account not only resource constraints in sensor nodes. but also the

Wireless Sewor Networks: ACopynght @ 2009 Institute

19

-

i:

NETWORK ARCHI

ll

1l

li

icrunes AND PRorocol srAcK

li

li

J this purpose, it is imPortant

flesign for WSNs.

fn network architectures and

br node structure and tYPicalldiscuss the classification of! a protocol stack for sensorrthis chapter.

I

lessi

I

ll

;.j.irrtrfle rtS of different network applications.

ir:rne a protocol stack to facilitate the

I-lis chapter introduces fundamental:, ",r'rol stack of WSNs. We wiII flrst introduce se., :\rlLrr JL<rw^ vr rr ur tr

,,-r.:trr n€twork architectures in Section 2.2, thei-i:.-or networks in Section2.3, and finally descrit' :: A L-rrks in Section 2.4. Section 2.5 will summariz

2.2 NETWORK ARCHITECTURES FOR WIR

sENSOR NETWORKS

.: lhis section, first we introduce the structure of a

:ipical network architectures for WSNs.

2.2.1 Sensor Node Structure

nsor node and then describe

\ scnsor node typically consists of four basic

-c.sinq unit, a communication unit, and a power it, which is shown in Fig.2.1'

,lr' sensing unit usually consists of one or!l phenomenon and generate

ifneaPCs convert the analogprocessing unit. The Process-

:\);rvcrters (ADCs). The sensors observe the phys

..:aioq signals based on the observed phenomenori.'""r[ into digital signals, which are then fed to th

:.r unit usually consists of a microcontroller or,..s... lntel's StrongARM microprocessor and I,ilrich provides intelligent control to the sensor I

L-rnsists of a short-range radio for performing

'rcr a radio channel. The power unit consists of

Power Unit

Processing Unit

Sensing Unit

nicroprocessor with memotymel's AVR microProcessor),

0

Fig. 2.1 Sensor node

nents: a sensing unit, a Pro-

and analog-to-digital

The communication unit. transmission and recePtionbattery for suPPlYing Power

i

NETWORK ARCHITECTURES LESS SENSOR NETWORKS

to drive all other in the system. In addition, a sensor node can alsobe equipped with some units, depending on specific applications. Forexample, a global positioni system (GPS) may be needed in some applications

ion for network operation. A motor may be neededthat require location i

21

2.2,2 Network Arch

A sensor network of a large number of sensor nodes denselvdeployed in a region of i , and one or more data sinks or base stations thatare located close to or i the sensing region, as shown in Fig.2.2.The sink(s)sends queries or comma

to move sensor nodes rna small module with low

2.3. However, long-distation. In sensor networks.

required transmission powision distance.Therefore. it ision distance in order to

sensing tasks. All these units should be biiilt intoconsumption and low production cost.

ransmission is costly in terms of energy consump-nergy consumed for communication is much higher

exponentially with the increase of transmis-to reduce the amount of traffic and transmis-

energy savings and prolong network lifetime.

Sensing region

node

sensor nodes collaborate toto the sink(s). Meanwhile,:

the sensor nodes in the sensing region while thermplish the sensing task and send the sensed datasink(s) also serves as a gateway to outside net-

works, for example, the Isimple processing on the

et. It collects data from the sensor nodeq performsdata, and then sends relevant information (or

the processed data) via thinformation.

rnet to the users who requested it or use the

To send data to the sin each sensor node can use single-hop long-distancetransmission, which leads single-hop network architecture, as shown in Fig.

than that for sensing and For example, the energy sonsumed fortransferring one bit of dati a receiver at 100m away is equal to that needed

]. The ratio of energy consumption for communi-to execute 3.000 instructcating 1 bit over the wi medium to that for processing the same bit couldbe in the range of 1,000- [3,4]. Furthermore, the energy consunied fortransmission dominates I energy consumed for communication and the

Sensor network architecture.

22 NETWORK ARCHITEC

Fig. 2.3 Single-hoP neovork

are close to each other, which makes it feasible to use

tion. In multihop communication, a sensor node transmi

the sink via one or more intermediate nodes' which can

sumption for communication. The architectglg- of.1 1organized into two types: flat and hierarchical [5l,which a

tnic sections (Sections 2.2.2.L and2.2.2.2).

For this purpose, multihop short-distance communication

rort tensoi networks, sensor nodes are densely deploye

PROTOCOL STACK

highly preferred. lnand neighbor nodeslistance communica-s sensed data toward

the energY con-network can be

described in the next

2.2.2.1 Flat Architecture. In aflatnetwork,each plays the same rolers Due to the largentifier to each node

in a sensor network. For this reason' data gathering is ally accomPlished bY

fiuery to all nodes inthat have the data

node communicatesas relays Frgute 2.4

2.2.2.2 Hierarchical Architedure, In a sensor nodesnd their data to the

itting the data to

rhe sink. A node with lower energy can be used to Ithe sensing task and

send the sensed data to its cluster head at short distance, a node with higherta from its cluster

process can not onlY

rcduce the energy consumption for communication, but balance traffic loadll sensor nodes have

in performing a sensing task and all sensor nodes are

number of sensor nodes, it is not feasible to assign a glob

using data-centric routing, where the data sink transmits

the iensing region via flooding and only the sensor no

matching the query will respond to the sink' Each

uith the sink via a multihop path and uses its peer

illustrates the typical architecture of a flat network'

are organized into clusters, where the cluster members

.-lusteiheads while the cluster heads serve as relays for I

energy can be selected as a cluster head to process

in.rnb"tt and transmit the processed data to the sink'

ancl improve scalability when the network size grows Si

rhe same transmission capability, clustering must be p y performed in

WIRELESS SENSOR NETWORKS

Sink

2.4 Flat network architecture.

@ Clusterhead

o cluster member

2.5 Single-hop clustering architecture'

:i,

order to balance thetion can beto the sink and i

The majorhow to organize tstrategies Accordicluster heads, aarchitecture or a

respectively [8].sensor network can

':

flc load aqlpng all sensor nodes. Moreon"r, s,u aggrega-

: cluster hiads to reduce the amount of datd'transmittedthe energy efficiency of the network [6]. fwith clustering is how to select the cluster heads and

rsters [7]. In this context, there are mariy clustering

the distance between the cluster members and their

organized into a single-tier clustering architecture or a

itecture. Figure 2.7 illustrates an example of*he msltitier

** ,i;

multitier clustering

24

cnteria, WSNs can be classified into differentng to different

NETWORK ,ARCHITECT AND PROTOCOL STACK

@ Clusterhead

O Cluster member

Fig. 2.5 Multihop clustering

Sink

(D Tier2 cluserhead

O Tier 1 clutcrhead

O Tier 0 cluster member

Fig. 2.7 Multitier clustering

clustering architecture [9].To address the clustering m, a variety of cluster-rng algorithms have been proposed in the literature [7- iThe reader is referredro Chapter 6 for a comprehensive introduction of these g algorithms.

2,3 CTASSIFICATIONS OF WIRELESS SENSOR KS

\\ SNs are application specific.A sensor network is usua deployed for a specific

arlplication.and thus has some different characte

CLASSIFICATIONS OF WIRE SENSOR NETWORKS

Static and Mobilesensor network ca

nodes are static wi

etvvork. According to the mobility of sensor nodes, a

static or mobile. In a static sensor network, all sensor

out movement, which is the case for many applications'

inistic sensor network, the positions of sensor nodes

25

However.somesensing task. A wi

example of mobile sensor networks [14]. Comparedtworks, whic.h is simpJer to co::trol and easier to imple-mobile sensor networks must consider the mobilityment, the design,

effect, which incri the cornplexity of implementation' ,

ardetsminktir Ne*vark According to the depioy-a sensor network can be deterministic or nondeter-

ior applications requite mobile nodes to at*onrplish aless bi,osensor network using autonomously controlled

animals is a typiciwifl staficsensorn

are preplanned a

be used in someare flxed once deployed. This type of network can onlyited situations, where the preplanned deployment is

possible.In most uations, however, it is difficult to depioy sensor nodes

in a preplannedi because of the harsh or hostile environments.

Instead. sensor are randomly deployed without preplanning and

engineering. Obvi nondeterministic networks are more scalable and

flexible, but requi higher control complexity.

. Static-Sink and M ite-Sink Network. A data sink in a sensor network can

be static or a static-sink network, the sink(s) is static with a fixedto or inside a sensing region. All sensor nodes sendposition located

their sensed data the sink(s). Obviously, a static sink makes the networksimpler to contro t it would cause the hotspot effect [5].The amount oftraffic that sensor are required to forward increases dramatically as

the distance to data sink becomes smaller. As a result, sensor nodes

nk tend to die early, thus resulting in network partitionnormal network operation. In a mobile-sink network,

in the sensing region to collect data from sensor

the traffic load of sensor nodes and alleviate the

ment of sensorministic. In a d

closest to the datl

hotspot effect in

Single-Sink andsink ormultiple s

and even disruptithe sink(s) movesnodes, which canl

network.

Network. A sensor network can haye a single

In a single-sink network,there is only one sink located

close to or inside Ito this sink. In a i

different positisend their data closest sink. which can effectivelv balance the trafficload of sensor n and alleviate the hotspot effect in the network.

. Single-Hop andbetween asensor

Network According to the number of hops

sensing resion.All sensor nodes send their sensed data

ttisink-neffiork, there may be several sinks located inlose to or inside the sensing region. Sensor nodes can

and the data sink, a sensor network can be classified

into single-hop multihop. In a single-hop network, all sensor nodes

transmit their sen data directly to the sink,which makes network controlsimpler to imple nt. However, this requires long-range wireless commu-nication. which is in terms of both etergy consumption and hardware

26 NETWORK ARCHITEC

implementation. The furthest nodes from thequickly than those close to the sink. Also. thenetwork may increase rapidly with the increase owould cause more collisions, and thus inereasedelivery latency. In a multihop network, sensordata to the sink using short-range wirelessintermediate nodes Each intermediate nodeforward the data along a multihop pa!S. Morecbe performed at an intermediate node [6 efiminacan reduce the total amount of traffic in the neenergy efficiency of the network. In general, asimpler network architecture and thus ii easier toapplications in small sensing areas with sparselyMultihop networks have a wider range of applicatcontrol complexity.S elf- Reconfigurab le and Non- Self-Configurable N rconfigurability of sensor nodes, a sensor networkor non-self-conligurable. In a non-self-configurabhave no ability to organize themselves into a netto rely on a central controller to control each sinformation from them. Therefore, this type offor small-scale networks In most sensor nit*o,are able to autonomous$ organize and maintathemselves and collaboratively accomplish a senssuch self-configurability is suitable for large_scacomplicated sensing tasks.

Homogeneous and Heterogeneous Network.nodes have the same capabilities, a sensoror heterogeneous [15]. In a homogeneous networthe same capabilities in terms of energy, computaltrast, a heterogeneous network has some sophistiare equipped with more processing and communnormal sensor nodes. In this case, the network canand communication tasks to those sophisticated ncits energy efficiency and thus prolong the lifetime.

2.4 PROTOCOL STACK FOR WIRELESS SENSOR

The protocol stack for WSNs consists of five protocol ladata link layer, network layer, transport layer, and applicFig. 2.8. The application layer contains a viriety of aigenerate various sensor network applications. The tifor reliable data delivery required o-y the application network layer is

AND PROTOCOT STACK

will die much moretraffic Ioad in the

network size, whichconsumption and

transmit their sensedion via one or more

perform routing anddata aggregation can

redundancy, whichand thus improve the

-hop network hasntrol.It is suitable forployed sensor nodes.

at the cost of higher

According to thebe self-configurabletwork, sensor nodes

ik. Instead, they havenode and collect

ks is only suitable', sensor nodes

;heir connectivity byitask. A network withbetworks to perform

ng to whether sensorbe homogeneous

ll sensor nodes have; and storage. In con-

sensor nodes thating capabilities than

more processingin order to improve

: the physical layer,Iayer, as shown in

Jayer protocols tolayer is responsible