er | IM OPEN CALL A Survey on Sensor Networks lan F. Akyildiz, Weilian Su, Yogesh Sankarasubramaniam, and Erdal Cayirci Georgia Institute of Technology ABSTRACT Recent advancement in wireless communica tions and electronics has enabled the develop- ment of low-cost sensor networks. The sensor networks can be used for various application areas (eg, health, military, home). For different application areas, there are different technical insues that researchers are currently resolving, The current sate ofthe art of sensor networks is captured inthis article, where solutions are dis- cussed under their related protocol stack layer sections. This article also points out the open research issues and intends to spark new inter~ ests and developments in this Fld INTRODUCTION Recent advances in wireless eommonicatios and electronics have enabled the development of lo cost, lowspower, multifunctional semsor nodes that are Small in size and communicate untethered in short distanees. These tiny sensor nodes, which consist of sensing, data processing, and communi- eating components, everage the idea of sensor networks. Sensor networks representa significant improvement over trtional sensor ‘A sensor network is composed of a large number of sensor aodes that are densely deployed either inside the phenomenon or very close to it, The position of sensor nodes need not be engineered or predetermined. This allows. Fandom deployment in inaccessible terrains or disaster relief operations, On the other hand this also means that sensor network protocols and algorithms must possess self-organizing capabilites. Another unique feature of sensor networks is the cooperative effort of sensor nodes. Sensor nodes are fitted with an onboard processor. instead of sending the raw data to the hodes responsible for the fusion, they use their processing abilities to locally carry out simple Computations and transmit oly the tequired and partially processed data “The above described features ensure a wide range of applications for sensor networks. Some of the application areas ate health military, and ample, the rapid deploy ment, sellorganization, and fault tolerance chat acteristics of sensor networks make them a very promising sensing technique for military com ‘mand, control, communicaiions, computing, inte gence, surveillance, reconnaissance, and targeting systems. In health, sensor nodes can also be deployed to monitor patients and assist disabled patients, Some other commercial applications Include managing inventory, monuoring product quality, and montoring disasier areas. Realization of these and other sensor nets work applications requize wireless ad hoc net working technigues, Although many protocols and algorithms have been proposed for tradi tional wireless ad hoc networks, they ate not wel suited to the unique features and applies ‘ion requirements of sensor networks. To illus trate this point, the differences between sensor networks and ad hoe networks are +The number of sensor nodes in sensor aet- work ean be several orders of magnitude higher than the nodes in an ad hoe network + Sensor nodes ace densely deployed fensor nodes are prone to failures, * The topology of a sensor network changes very frequently. + Sensor nodes mainly use a broadeast com- ‘munication paradigm, whereas most ad hoc networks are based on point-to-point com + Sensor nodes are limited in power, compu tational capacities, and memory + Sensor nodes may not have global kdentiew ton (ID) because of the large amount of overhead and large number of sensors, Many researchers are currently engaged in devel= oping schemes that fulfil these requirements Tn this article we presenta survey of protocols, and algorithms proposed thus fa for sensor net- ‘works. Our aim is to provide a better understand- ing of the current research isues in this emerging field, We also attempt an investigation into per taining design constraints and outline the use of certain tools to meet the design objectives. “The remainder of the article is organized as follows, We discuss the communication architec- ture of the sensor networks as well a the factors that influence sensor network design. We pro: Vide a detailed investigation of eurrent proposals inthe physica, data link, network, transport, and application lavers, respectively. We then cons ‘ude our article 102 o163-6s090817.00 6 2002 EEE TEE Communications Magazine» August 2002‘SENSOR NETWORKS ‘COMMUNICATION ARCHITECTURE The sensor nodes ate usually scattered in a sensor {field as shown in Fig 1. Each ofthese scattered sensor nodes haste capabilities 1 collet data and route data back to the sink. Data are routed ‘pack to the sink by a multihop infrastructureess architecture through the sink as shows in Fig ‘The sink may communicate with the ask manager rode via Internet of satelite. The desig of the Sensor network as described by Fig, 1s intlie fenced by many factors, including faut tolerance, scalability, production costs, pers environment, Sensor network topology, hardware constrains, ‘nansmision media, a0 poworconstonpion. 1 DESIGN FACroRS The design factors are addressed by many researchers as surveyed inthis article. However, none of these studies has a fully integrated view ofall the factors driving the design of seusor networks and sensor nodes, These factors ate {Important because they serve as a guideline to ‘design a protocol or so algocthm for sensor net ‘works. in addition, these influencing factors can bbe used to compare diferent schemes, Fault Tolerance — Some sensor nodes may fi ‘or be blocked due to lack of power, or have physical damage or environmental interference ‘The failure of sensor nodes shoul not affect the ‘overall {28k ofthe sensor nctwork, This Is the reliability or fault tolerance issue. Fault toler ance is the abifigy to sustain sensor network functionalities without aay inteeruption due to sengor node tailures [1 2) The reliability (0) fr faul tolerance ofa sensor node fs modeled i [2 using the Poisson distribution to capture probability of not having a failure within the tine interval (04): Rul =e, 0 here 2g isthe failure rate of sensor node k and tithe time period Scalability — The number of sensor nodes deployed in studying a phenomenon may be on the ofder of hundreds or thousands. Depending fon the application, the number may reach an extreme Value of millios. New schemes must be able to work with this number of nodes. They ‘must also utilize the high density of the sensor networks: The density can range from few sensor nodes to few hundred sensor nodes in a region, Which caa be less than 10 m in diameter. The ‘ensity jean be calculated according to [3] a5 HR) = (VR RBVA, @ ‘where Wis the number of scattered sensor nodes in region A, and R is the radio transmission range. Basically, u(R) gives the aumber of nodes Within the transmission radius of each node in region A Production Costs — Since sensor networks ‘consist of a large number of sensor nodes, the bast of a single node is very important to justify the overall cost of the network. If the Gost ofthe __ has to be kept low. The state-0-the-a Intern and ‘orale f Sensor fld Sensor nodes Figure 1. Sensor nodes scanored na sensor fold Location finding system Mobiizer rocmses | =] | Jest{ Processor sano |ane Ese ae] : Figure 2. The components ofa sensor node network is more expensive than deploying tadi- ‘ional sensors, the sensor network is ot costjus- tified. As a result, the cost of each sensor node technol ‘ogy allows a Bluetooth radio system to be Tess than USS10 [4]- also, the price of a piconode is targeted to be less than USSI, The cast of sen sor node should be much less than USS1 in ‘order for the sensor network to be feasible. The ‘ost of Bluetooth radio, which is known to be 2 low-cost device, is even 10 times more expen- ‘ive than the targeted price for a sensor node. Hardware Constraints — A sensor node is ‘made up of four Basie components, as shown in Fig. 2 a sensing unit, a processing unit, a ranscio- erunit, and a power unit. They may also have ‘additional application-dependent components Such asa location finding system, power generator, ‘and mobilize, Sensing units are usually Composed ‘of we subunits sensors and analogto-igitt con: verters (ADCS). The analog signals produced by the sensors based on the observed phenomenon are converted to digital signals by the ADC, and then fed into the processing unit. The processing unit, which is generally associated with a small storage unit, manages the procedures that make the sensor node collaborate with the other nodes to carry out the assigned sensing tasks. A transooiver unt connect the node to the network ‘One ofthe most important eomponents ofa sen sor node is the power unit, Power units may be Supported by power seavenging units such as solar cel, Thete are also other subunits that are {IEEE Contmunications Magazine * Aug 2002 103— In a mutti-hop ‘sensor network, ‘communicating nodes are linked! by a wireless: medium. These links can be formed by radko, Infrared or optical media. To enable global operation of these networks, the chosen transmission medium must be available worldwide, application-dependent, Most ofthe sensor net- Work routing techniques and sensing tasks require knowledge of location with high acura: cy. Thus, itis common that a sensor node has @ Tocation Finding system. A mobilizer may some- times he needed 0 mave sensor nodes when itis required to carry out the assigned tasks ‘All of these subunits may nced to fit into @ matchbox-sized module [5}. The required size ‘may be smaller than even a cubic centimeter [6] Which is light enough to remain suspended in the air. Apart from size, there are tome ather Si font constraints for sensor nodes, These nodes ‘ust [7] consume extremely low power, operate in high volumetric densities, have low production cast, be dispensable and autonomous, operate ‘unattended, and be adaptive tothe environment Sensor Network Topology — Hundreds to several thousands of nodes are deployed throughout the sensor field. They ate deployed within tens of feet of each other [5], The node densities may be as high as 20 nodes/m’ [8] Deploying # high number of nodes densely requires careful handling of topology mainte rnance. We examine issues related (o topology ‘maintenance and change in three phases: *Predeployment and deployment phase: Sensor ‘nodes can be either theawn in 3s a mass oF phiced one by one in the sensor field. They Ean be deployed by dropping from a plane, deliveried in an artillery shell, rocket, oF missile, and placed one by one by either a human ora robot. Post-deployment phase: After deployment, topology changes are duc to change in sensor nodes’ [5] postion, reachability (duc to jam- ming, noise, moving obstacles, etc), available energy, malfunerioning, and tak details. Redeployment of additional nodes phase: ‘Additional sensor nodes exn be redeployed at any time to replace malfunctioning nodes (due to changes in task dynamic, Environment — Sensor nodes are densely deployed either very close or directly inside the phenomenon to be observed. Therefore, they ‘usually work unattended in remote geographic freas. They may be working in the interior of large machinery, atthe bottom of an ocean, in biologically or chemically contaminated fei, in a battlefield beyond the enemy lines, and in & hhome or large building Transmission Media — In a multihop sensor network, communicating nodes are linked by @ wireless medium, These links can be formed by rao, infrared, or optical media. To enable glob al operation of these networks, the chosen tuns- mission medium must be available worldwide, ‘Much of the current hardwaze for sensor nodes Js based on RF circuit design. The WAMPS wireless sensor node deserbed in [8] uses a Bluetooth-com patible 24 GH tanseciver with an integrated fr ‘quency synthesizer. The low-power sensor device ‘descr in [9 uss single-channel RF transce fer operating at 916 MHz, The Wireless Invgrated Nenwork Sensors (WINS) architecture [6] also uses radio Tinks for communication ‘Another possible mode of internode commu: nication in sensor networks is by infrared Infrared communication is license-free and robust to interference from electtieal devices. Tnfrared-hased transceivers are cheaper and eas ier to build. Another interesting development is that of the Smart Dust mote [7}, which is am autonomous sensing, computing, and commun tation system that uses the optical medium for transmission. Both infrared and optical require & Tine of sight between the sender and receiver Power Consumption — The wireless sensor ‘ode, being a microelectronic devie, can only te equipped wth a limited power source (< 03 Ai, LEV). In some spplieation sceneries, feplesishment of power resources might he impossible, Sensor node lifetime, therefore shows a strong dopendenee on battery lifetime inautnop fe stor nertork ach de ay the dua role of date originator and Gata fouter: The mafunetoning of afew nodes eon fase sgnican® topological changes and might ‘equire rerouting of packets and reorganization ofthe network. Henge, power consesation snd poser management take on additional iipor- {enoe. Is for these reas thal researchers are curently focusing onthe design of power-aware protools and algorithms for sensor networks The maia task of «sensor aode ita sensor Ged is to detect event perform quick local data processing, and then transmit the Gia, Powe oo Sumption can hence be divided int three domains Sensing commision, an data proces Prorocot STACK ‘The protocol stack used by the sink and sensor nodes shown in Fig. 1s given in Pig, 3. This pro- tocol stack combines power and routing aware- ness, integrates data with networking protocols, communicates power efficiently through the wire Jess medium, and prometes cooperative cifors of sensor nodes. The protocol stack consists of the phosical layer, data lnk layer, network layer, tras por layer, application laser, power management plane, mobility management plane, and task man fagement plane, The physical layer addresses the needs of simple but robust modalation, transmis sion, and receiving techniques Since the environ iment is noisy and sensor nodes can be mobile, the medium 3ecess control (MAC) protocol must be pawer-aware and able to minimize collision with neighbors’ broadcasts. The network layer takes cafe of routing the data supplied by the transport layer. The transport layer helps (0 ‘maintain the Mow of data if the sensor networks application requires it. Depending on the sensing tasks, different types of application software can bee built and used on the application layer. In addition, the power, mobility, and task manage- ‘ment planes monitor the power, movement, and {ask distribution among the sensor nodes. These planes help the sensor nodes coordinate the sens- dng task at lower overall power coasumption. "The power management plane manages how a sensor node uses its power. For example, the ‘sensor node may turn off its receiver after reeeiv« ing a message from one ofits neighbors. This is to avoid getiing duplicated messages, Also, when the power level of the sensor node is low, the sensor node broadcasts to its neighbors tha iis 104 {EEE Communications Magazine + August 2002