November 2011
, published 28, doi: 10.1098/rsta.2011.0382370 2012 Phil. Trans. R. Soc. A Niki Trigoni and Bhaskar Krishnamachari
Sensor network algorithms and applications References
alsocietypublishing/content/370/1958/5.full.ht This article cites 10 articles, 9 of which can be accesd free Subject collections (68 articles)
全国四六级成绩查询
computer modelling and simulation collections Articles on similar topics can be found in the following Email alerting rvice
here in the box at the top right-hand corner of the article or click Receive free email alerts when new articles cite this article - sign up
Phil.Trans.R.Soc.A(2012)370,5–10
doi:10.1098/rsta.2011.0382
I N TRO D UCT I O N
Sensor network algorithms and applications
B Y N IKI T RIGONI1AND B HASKAR K RISHNAMACHARI2,*
1Department of Computer Science,University of Oxford,Wolfson Building,
suede
Parks Road,Oxford OX13QD,UK
2Viterbi School of Engineering,University of Southern California,Los Angeles,
CA90089,USA
A nsor network is a collection of nodes with processing,communication and nsing capabilities deployed in an area of interest to perform a monitoring task.There has now been about a decade of very active rearch in the area of nsor networks,with significant accomplishments made in terms
cartier是什么意思of both designing novel algorithms and building exciting new nsing applications.This Theme Issue provides a broad sampling of the central challenges and the contributions that have been made towards addressing the challenges in thefield,and illustrates the pervasive and central role of nsor networks in monitoring human activities and the environment.
2020开学时间>如何快速美白皮肤occupy是什么意思Keywords:networks;nsors;wireless;algorithms;protocols;applications
1.Introduction
The Issue is organized into three main themes:thefirst theme focus on the problem of network configuration and software,and introduces algorithms for node scheduling,time synchronization and energy management,as well as operating systems and network protocols.The cond theme covers a range of signal-processing techniques developed for nsor node localization,event detection and tracking,as well as in-network data compression.The third theme provides a taste of nsor network applications and covers a variety of t-ups,from monitoring controlled building environments to nsing challenging underwater environments,and from traditionalfixed nsor topologies to new domains of mobile participatory nsing.
2.Network configuration and software
In a typical nsor network,nodes can collect data from local nsors,perform local computations and communicate wirelessly with other nodes by radio.Given the limited radio communication range,each node can directly exchange messages *Author for correspondence(bkrishna@usc.edu).
One contribution of11to a Theme Issue‘Sensor network algorithms and applications’.
6N.Trigoni and B.Krishnamachari
with a limited number of neighbouring nodes.Thus,a nsor network can be viewed as a distributed system,in which nodes must carefully coordinate their communication and computation tasks in order to efficiently achieve a monitoring goal.Lenzen&Wattenhofer[1]highlight the power of distributed algorithms in solving network coordination tasks,such as carefully scheduling radio transmissions to avoid interference and synchronizing the clocks of nsor nodes.They highlight important differences between a wireless nsor network and an idealized distributed system,and discuss how the differences affect the design of distributed algorithms and their complexity.They show that the underlying communication model,which governs how nodes communicate and interfere with each other,is key to designing distributed algorithms for nsor networks,and has sparked new interest in the area of distributed computing.
The fact that nsor nodes are embedded within a physical space gives ri to interesting geometrical structures,and motivates the study of a special class of algorithms,referred to as geometric algorithms.Gao&Guibas[2]identify four distinct areas in which geometric structures can be exploited to configure nsor network operation.First,they can be ud in network localization, infer the locations of nsor nodes from local measurements,such as distance and angle estimations between neighbouring nodes.The authors give an overview of distributed and centralized localization algorithms,explain the challenge of localization ambiguity(multiple localization solutions that satisfy the distance/angle constraints)and summarize key theoretical results on graph rigidity and network localizability.Second,geometrical information can be ud in routing.The authors discuss how physical and virtual coordinates of nsor nodes can be ud to determine the path that a packet follows to a destination.They also overview routing techniques that aim to balance the network communication load,as well as landmark-bad routing schemes in which a small subt of nodes is lected as the landmark and routing decisions take into account distances to landmarks.Third,geometrical information can be exploited to design information brokerage mechanisms,which enable information consumers and information sources to discover each other in a communication-efficient manner.Finally,the authors highlight the u of geometrical algorithms in discovering and reprenting non-trivial network topologies with‘hole’structures.
A key aspect of designing network configuration algorithms is to conrve energy and prolong the lifetime of the network.Stankovic&He[3]survey the area, and prent three distinct approaches to energy management.Thefirst approach focus on the hardware side,and highlights ways of scavenging and storing energy at the nsor nodes.The authors also discuss differences in energy consumption in various types of nsor hardware,and controls over hardware components to allow trade-offs between energy and performance.The cond approach,referred to as energy management in the small,concerns energy management solutions that are targeted at specific layers of nsor node functionality,for example medium access control,routing,localization and time synchronization.The third approach,referred to as energy management in the large,compris efforts to build system components that are specifically dedicated to energy management. Such components typically orchestrate nsor activation patterns,taking into account the nsing range and density of nsor nodes,as well as the nature of the monitored heir duration,shape,mobility,and so on).
Introduction7 Network protocols define the overall communication process in a nsor network,and require an operating system at each node to provide a software abstraction to manage its resources.Dunkels&Dutta[4]provide an overview of developments in the design of operating systems
and network protocols for nsor networks.Focusingfirst on operating systems,the authors identify challenges pertaining to resource constraints and the sheer diversity of both hardware and applications,and discuss how the challenges have been addresd in the literature.They highlight the design issues pertaining to concurrency and execution models,memory allocation,storage,energy and the communication architecture provided by nsor network operating systems.Then,they prent network protocols for low-power medium access,routing and reliable transfer.As nsor networks mature and transition to industrial practice,there is an incread effort on the standardization of network protocols.The authors describe some of the ongoing efforts by the Internet Engineering Task Force to enable Internet protocol-bad nsor networks.
3.Signal processing
diantaiOnce nsors lf-configure into a network,they must t off to address the monitoring task at hand.This involves activating their nsors,and processing nsor signals to infer events of interest.The events are then stored locally or communicated wirelessly through the network.Signal processing has a number of interesting applications,from localizing nsor nodes to detecting and tracking events of interest and compressing data within the network prior to data propagation.
英讯理想Lédeczi&Maróti[5]provide an interesting overview of range-bad localization methods for wireless nsor networks.They review various approaches to determining the distance between two nodes,by processing radio,acoustic or light signals.They discuss the advantages and disadvantages of ranging using time offlight versus signal strength and pha measurements.The authors then discuss how the collected ranging data and their error distributions can be fud to provide location estimates.They also show how localization techniques can exploit a priori information about the location of some nodes,the environment where the network is deployed and patterns of mobility(in scenarios of localizing mobile nodes).While global positioning system(GPS)technology has largely addresd the problem of node localization in outdoor environments,there is a large number of interesting rearch challenges that need to be tackled both indoor and in den urban environments.The authors believe that some of the challenges may be overcome by multi-modal localization,that is,by fusing signals from multiple nsors that measure different physical phenomena(including GPS, cameras,inertial nsors,air pressure nsors,and so on).
Localization is only one instance of a large class of inference problems in the context of nsor networks.Veeravalli&Varshney[6]summarize rearch developments in the area of distributed inference,with emphasis on dealing with resource constraints,such as limited power,communication r
ange and bandwidth.They discuss three different types of inference problems:distributed detection of events of he prence of a contaminant),parameter he location and intensity of a heat source),and trackingbelievein
8N.Trigoni and B.Krishnamachari
(e.g.work-in-process tracking in a manufacturing environment).The authors point out that,although significant contributions have already been made in addressing the problems,there remain a lot of open rearch issues.Fertile areas for future work include distributed inference with conditionally dependent obrvations,when the obrvation statistics are unknown or partially known or when obrvations are made by nsors of different modalities.
Once nsor data are gathered and procesd locally at nsor nodes,they typically need to be transmitted wirelessly through the network to special purpo nodes,called bastations or sinks,to be consumed by the application urs. Compressing data as they are transmitted to the sink can significantly reduce the network communication load and increa the lifetime of the network.In their article,Duarte et al.[7]review a number of distributed compression techniques for wireless nsor networks.They pay particular attention to spatial compression techniques,which exploit correlations i
tdscdma是什么意思n data generated by neighbouring nodes. They make the key obrvation that such techniques require raw data to be transmitted from the source nodes to aggregating nodes,where spatial correlations are discovered and data are compresd.Existing algorithms vary in the degree of synergy between the coding and communication aspects:from treating routing and compression independently to prioritizing one over the other, or jointly optimizing the two aspects.The importance of distributed compression is likely to increa over the next years,as nsor networks become more ubiquitous,increa in size,density and sampling rates,and bandwidth becomes a scarce resource.
4.Applications and new domains
In the last decade,we have en significant advances in hardware and algori-thm design for nsor networks.Together,they have enabled the adoption of this exciting new technology in a variety of interesting applications,from indoor to outdoor nsing,from monitoring human activities to nsing large-scale environmental process.Owing to space reasons,it is infeasible to provide an exhaustive list of nsor network applications and their specific rearch challenges.To illustrate the breadth of existing applications,we include three examples with completely different application requirements and network t-ups.
The article by Liu&Terzis[8]describes the unique challenges of using nsor networks to monitor data centres.Data centres can vary in size,from a few rver cabinets to large dedicated buildings consuming hundreds of megawatts of electricity.Careful monitoring of large data centres can lead to significant energy savings,and lower maintenance costs.Different rvers run different rvices, and the number of urs that access the rvices may vary significantly over time.Therefore,the physical locations of rvers,the kinds of rvices they run and the time-varying load incurred by each rvice have an impact on the stress put on the power and cooling systems.Fixed nsor nodes are an attractive method for increasing visibility into a data centre’s operating conditions.They can sample environmental attributes at high temporal and spatial resolutions, notify operators of unsafe operating conditions,trace the source of failures and help control the power and cooling systems.
