An analysis of a large scale habitat monitoring application.pdf

《An analysis of a large scale habitat monitoring application.pdf》由会员分享，可在线阅读，更多相关《An analysis of a large scale habitat monitoring application.pdf（13页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、AnAn AnalysisAnalysis ofof a a LargeLarge ScaleScale HabitatHabitat MonitoringMonitoringApplicationApplicationRobert Szewczyk, Alan Mainwaring , Joseph Polastre, John Andersonand David CullerEECS DepartmentIntel Research BerkeleyCollege of the AtlanticUniversity of California, Berkeley2150 Shattuck

2、Avenue105 Eden St.Berkeley,California 94720Berkeley,California 94704Bar Harbor, ME 04609ABSTRACTABSTRACTHabitat and environmental monitoring is a driving application forwireless sensor networks. We present an analysis of data from asecond generation sensor networks deployed during the summerand autu

3、mn of 2003. During a 4 month deployment, these net-works, consisting of 150 devices, produced unique datasets forboth systems and biological analysis. This paper focuses on nodaland network performance, with an emphasis on lifetime, reliabil-ity, and the the static and dynamic aspects of single and

4、multi-hopnetworks. Wecompare the results collected to expectations set dur-ing the design phase: we were able to accurately predict lifetime ofthe single-hop network, butwe underestimated the impact of multi-hop traffic overhearing and the nuances of power source selection.While initial packet loss

5、data was commensurate with lab experi-ments, over the duration of the deployment, reliability of the back-end infrastructure and the transit network had a dominant impacton overall network performance. Finally, we evaluate the physicaldesign of the sensor node based on deployment experience and apos

6、t mortem analysis. The results shed light on a number of de-sign issues from network deployment, through selection of powersources to optimizations of routing decisions.CategoriesCategories andand SubjectSubject DescriptorsDescriptorsC.2.1 Computer-CommunicationComputer-Communication Networks:Networ

7、ks: Network Archi-tectureandDesignWirelessCommunications; C.3Special-PurposeSpecial-PurposeAndAnd Application-BasedApplication-Based SystemsSystems: Real-Time and embedded sys-tems; C.4 PerformancePerformance ofof SystemsSystems: Design StudiesGeneralGeneral TermsTermsPerformance, Design, Implementa

8、tionKeywordsKeywordsSensor Networks, Habitat Monitoring, Microclimate Monitoring,Network Architecture, Long-Lived Systems, Application Analysis1. 1.INTRODUCTIONINTRODUCTIONPermission to make digital or hard copies of all or part of this work forpersonal or classroom use is granted without fee provid

9、ed that copies arenot made or distributed for profit or commercial advantage and that copiesbear this notice and the full citation on the first page. Tocopy otherwise, torepublish, to post on servers or to redistribute to lists, requires prior specificpermission and/or a fee.SenSys04, November 35, 2

10、004, Baltimore, Maryland, USA.Copyright 2004 ACM 1-58113-879-2/04/0011 .$5.00.A broad class of applications are within the reach of contempo-rary wireless sensor networks (WSNs). These applications sharea common structure, where fields of sensors are tasked to take pe-riodic readings, and report res

11、ults and derived values to a centralrepository. There are both scientific and commercial applications,for example: microclimate monitoring, plant physiology, animalbehavior 16, precision agriculture 2, 4, structural monitoring5 and condition-based maintenance. These sense-and-send appli-cations have

12、widely-varyingsampling rates and networkbandwidthdemands.In the context of habitat and environmental monitoring, WSNsoffer significant advantages. Individual devices can be made suf-ficiently numerous to take measurements at many locations of in-terest, and mitigate errors arising from the interpola

13、tion and ex-trapolation from coarser-grained samples. They can be sufficientlysmall to be co-located with phenomena of interest without alteringthe parameters to be measured. And they can be unobtrusively em-beddedintheenvironmentwithoutcreatingconspicuouslandmarksthat change the behaviors of its in

14、habitants.Long-term unattended operation enables measurement at spatialand temporal scales impractical with human observers or sparselydeployedinstruments. The lifetimes made possible with contempo-rary low-powermicroelectronics can prolong the duration of exper-imental observations. At the same tim

15、e, automation improves thedata quality and uniformity of measurement, while reducing datacollection costs as compared with traditional human-centric meth-ods. Devices can operate for prolonged periods in habitats that areinhospitable, challenging or ecologically too sensitive for humanvisitation. Un

16、obtrusiveobservation is key for studying natural phe-nomena.WSNs offer more capabilities than standalone dataloggers andwired instrumentation. Wireless telemetry is valuable because itminimizes observereffects,study site intrusions and environmentalalterations. Forexample, visits to study areas to m

17、onitor and down-load loggers are no longer necessary,while health and status of in-strumentation can be monitored remotely. More general network-ing offers great benefits, such as continuously updated databases ofsensor readings accessible through the web, access to live readingsfrom individual sens

18、ors, and is key to distributed in-network pro-cessing. These capabilities may yield new experimental designs,and paradigms for data publication, dissemination, and scientificcollaboration.We have incrementally deployed several sensor networks of in-creasing scale and physical extent in a wildlife pr

19、eserve. Whileamassing a novel dataset for biological analysis, the annotated dataare interesting from a systems perspective. The packet logs froma single-hop and multi-hop network reveal insight on lifetimes,packet yields, network structure and routing. For example, someFigureFigure 1:1: GeospatialG

20、eospatial distributiondistribution ofof petrelspetrels obtainedobtained byby directdirect humanhuman observationobservation (left)(left) andand a a particularparticular featurefeature ofof thethe habitathabitat(average(average temperaturetemperature atat midnightmidnight inin thethe burrowsburrows (

21、center)(center) andand onon thethe surfacesurface (right)(right) collectedcollected fromfrom outout sensorsensor network)network)nodes ran for nearly four months but some for just a few days.Analysisrevealschangesinnetworkstructureandperformanceoverthe lifetime of the deployment. Though the applicat

22、ion was sim-ple, it exhibited interesting and unexpected behaviors after its ini-tial setup. Although it is representative of applications with lowsampling and bandwidth demands, its architecture and implemen-tation are general and thus provides a reference point for others inthis space.The remainde

23、r of this paper presents an analysis of that data col-lected during the summer and autumn of 2003 from two sensornetwork deployments on Great Duck Island, Maine. It is organizedas follows: Section 2 describes the application, system architec-ture and realization. Section 3 is an analysis of the data

24、. Section 4presents experiences and lessons learned. Section 5 discusses re-lated works and Section 6 concludes.micro-climate data. The first GIS plot shows predicted populationdensity on the island based upon direct inspection of burrow occu-pancyfrom an entire season of sampling - months of labor

25、resultingin a single plot. The latter two plots show temperatures in the un-derground burrows and at the corresponding points on the surface.Data for these was collected by our sensor network at midnight ona typical summer evening. Darker colors are warmer temperatures,lighter colors correspond to c

26、ooler temperatures, and and shadedarea of similar colors are isoplethic temperature regions.Cooling surface temperatures are apparent, whereas the buffer-ing and insulating properties of the burrowscause them to maintaina nearly constant temperature. Hot-spots in underground burrowsare of special in

27、terest. For these hot-spots there is mounting evi-dence from direct inspection and acoustic playbacks that a residentpetrel produces the heat. Once the correlation between warmer bur-row temperatures and occupancy can be definitively established,expected population density visualizations could be re

28、placed bynightly, or even hourly, sensor data. This would represent a fun-damental advancement towards the understanding the distributionand abundance of this species.2. 2.SYSTEMSYSTEMAnticipating to the analyses in Section 3, this section presentsbackground on the application, the system architectu

29、re and its im-plementation. In particular,it describes the tiered network architec-ture typical of habitat monitoring applications and the design andimplementation of its core components.2.22.2ArchitectureArchitecture2.12.1ApplicationApplication BackgroundBackgroundJohn Anderson was studying the dis

30、tribution and abundance ofsea birds on an offshore breading colony on Great Duck Island,Maine. Hewantedtomeasuretheoccupancyofsmall, undergroundnestingburrows,andtheroleofmicro-climaticfactorsintheirhabi-tat selection. Wehypothesized that a sensor network of motes withappropriate sensors, with TinyO

31、S components for low-power rout-ing and operation, could log readings in a web-accessible database.It seemed plausible that passive infrared (PIR) sensors could di-rectly measure heat from a seabird, or that temperature/humiditysensors could measure variations in ambient conditions resultingfrom pro

32、longed occupancy. We also wanted, in a modest way,to translate the vision for sensor networks to a concrete reality.This simple application would require creating a complete hard-ware/software platform, firmly grounded in the needs of a tradi-tional ecological study. It emphasized small mote size, l

33、ong life-time, unattended operation, and caused us to consider the verifica-tion and ground-truth of sensor readings.Among the life scientists, Graphical Information Systems (GIS)have become the lingua franca for the visual presentation, analysisand exchange of geospatial data. Weimagined a system c

34、apable ofproducing animal density GIS plots like Figure 1, and at the con-clusion of 2003, the system could generate such visualizations forThe system we deployed has the tiered architecture shown inFigure 2. The lowest end consists of sensor nodes that performcommunication, computation and sensing.

35、 They are typically de-ployedin sensor patches. Depending on the application, theymightform a linear transect, a grid region, or a volume of nodes nodes forthree-dimensional monitoring. Each sensor patch has a gatewaythat sends data from the patch through a transitnetwork to a remotebase station via

36、 the base station gateway. We expect mobile fieldtools will allow on-site users to interact with the base station andsensor nodes to aid the deployment, debugging and management ofthe installation. The base station provides Internet connectivityanddatabase services. It should handle disconnected ope

37、ration fromthe Internet. Remote management facilities are a crucial featureof a base station. Typically the sensor data is replicated off-site.These replicas are located whereverit is convenientfor the users ofthe system. In this formulation, a sensor network consists of oneor more sensor patches sp

38、anned by a common transit network andbase station.Sensor nodes are small, battery-powereddevices capable of gen-eral purpose computation, bi-directional wireless communication,and application-specific sensing. The sizes of nodes are commen-surate with the scale of the phenomenon to be measured. Thei

39、r life-time varies with duty cycle and sensor power consumption; it canbe months or years. They use analog and digital sensors to sam-Sensor PatchesVerificationNetworkGatewaySingle Hop NetworkTransit NetworkGatewayClient Data BrowsingGatewayBase stationMulti Hop Networkand ProcessingSite WAN LinkInt

40、ernetFigureFigure 2:2: ArchitectureArchitecture ofof thethe habitathabitat monitoringmonitoring systemsystemple their environment, and perform basic signal processing, e.g.,thresholding and filtering. Nodes communicate with other nodeseither directly or indirectly by routing through other nodes.Inde

41、pendent verification networks collect baseline data from ref-erence instruments that are used for sensor calibration, data valida-tion, and establishing ground truth. Typicallyverification networksutilize conventional technologies and are limited in extent due tocost, power consumption and other dep

42、loyment constraints.2.32.3ImplementationImplementationThe deployment was a concrete realization of the general archi-tecture from Figure 2. The sensor node platform was a Mica2Dot,a repackaged Mica2 mote produced by Crossbow, with a 1 inchdiameter form factor. The mote used an Atmel ATmega128 mi-cro

43、controller running at 4 MHz, a 433 MHz radio from Chipconoperating at 40Kbps, and 512KB of flash memory. The mote inter-faced to sensors digitally using I2C and SPI serial protocols and toanalog sensors using the on-board ADC. The small diameter circuitboards allowed a cylindrical assembly where whe

44、re sensor boardswere end caps with mote and battery internal. Sensors could beexposed on the end caps; internal components could be protectedby O-rings and conformal coatings. This established a mechanicaldesign where sensor board and battery diameters were in the 1 to1.5 inch range but the height o

45、f the assembly could vary.We designed two different motes for the application, burrowmotesfordetectingoccupancyusingnon-contactinfraredthermopilesand temperature/humidity sensors and weather motes for monitor-ing surface microclimates. The burrow mote had to be extremelysmall to be deployed unobtrus

46、ively in nests typically only a fewcentimeters wide. Batteries with lithium chemistries were chosenFigureFigure3:3: MoteMote configurationsconfigurations usedused inin thethe deployment:deployment: weatherweathermotemote (left)(left) andand burrowburrow motemote (right)(right)because discharge volta

47、ges remained in tolerance for mote, radioand sensors almost to the end of the battery lifetime. We electednot to use a DC boost converterbecause it introduces noise and in-creases power consumption. The motes operation and the qualityof its sensor readings depends upon the voltage remaining withinto

48、lerance. Should the voltage fall outside the operating range of themote, radio, or sensors, the results are unpredictable.2.3.1Burrow and WeatherMotesBurrow motes monitor temperature, humidity and occupancy ofnestingburrowsusingnon-contactpassiveinfraredtemperaturesen-sors. They have two sensors: a

49、Melexis MLX90601 non-contacttemperature module and a Sensirion SHT11 temperature and hu-midity sensor. The Melexis measures both ambient temperature(1 C) and object temperature (2 C). The Sensirion measuresrelative humidity (3.5% but typically much less) and ambienttemperature (0.5 C), which is used

50、 internally for temperaturecompensation. The motes used 3.6V Electrochem SB880 batteriesrated at 1Ahr, with a 1mA rated discharge current and a maximumdischarge of 10mA. The 25.4mm diameter by 7.54mm tall dimen-sions of the cell were well suited for the severely size constrainedburrow enclosures.Wea