wireless sensor networks - unibo.itlia.deis.unibo.it/courses/retils/retils-0304/seminari/wsn.pdf ·...
TRANSCRIPT
1
A Survey onWireless Sensor Networks
Eugenio Magistretti
Bologna, February 12th 2004
Reti di Calcolatori LS
2
Table of Contents
• WSN Concepts:– WSN characteristics– Why WSN?– WSN architecture and design guidelines
• Solutions for networking support:– Adaptive Topology: GAF– Data Dissemination: LEACH, Pegasis, Teen, SPIN– In-network processing: Directed-diffusion– User Queries and Database Oriented Approach:
COUGAR– Conclusions and On-going Work
3
PART 1
WSN Concepts
4
Goals…• Link Physical Word and Digital Data Networks,
providing distributed network and Internet access tosensors, controls, processors deeply embedded in equipment, facilities and environment
• Technical goals: scalability, lifetime, adaptability
• Huge number of participants (billions)
• Generally composed by randomly placed and stationary devices
• Limited resources (e.g. battery-powered devices) • Low bit rate delivery to the user • Short message packets suffice
…and features
5
Devices• UCB PC104 based
– AMD ElanSC400 CPU
– SDRAM 16MB
– Flash Disk 16MB
– Radio Packet Controller 418 MHz
– Use motes as radio
– Linux OS
• Motes: Smart-It– Atmel ATmega 128L (8 MHz 8 MIPS)– 64Kbyte RAM, 128Kbyte FLASH ROM,
4Kbyte EEPROM– Bluetooth radio (57.6 kbps)– Analog in: 8 x 10-bit AD converter – Digital IO: 16-bit – Interrupt lines: 3, edge or level triggered – Serial IO at 57.6Kbps– Tiny OS
6
Applications
• Ubiquitous and pervasive computing• Environmental monitoring (air, water, soil
chemistry; surveillance)
7
Applications
• Home automation (smart houses, virtualneighbor)
• Inventory tracking (in warehouses, laboratories)
• Futuristic: – Health monitoring (ingested sensors, smart
medications)Circulatory Net
8
Comparison to ConventionalNetworks
•Topology is a star network where a master has up to seven slaves(piconet); there are mechanisms to form a multihop topology
•Nodes are appliances and electronic consumer devices
•Nodes are short-range mobile
•Energy isn’t generally an issue
•Goal is to replace cable between devices and provide RF connection between them
Bluetooth
•Tens to hundreds of nodes
•Nodes are appliances outfitted with sophisticated radio transceivers
•Nodes are fully mobile
•Energy consumption is of secondary importance
•Aims to form and maintain a connected multihop network
•The goal is to provide QoS (throughput vs. delay)
MANET
•Thousands of nodes
•Nodes are appliances outfitted with sophisticated radio transceivers
•Mobile nodes greatly outnumber stationary (BS)
•BS have unlimited power supply, mobiles are battery-operated
•The primary goal is to provide high QoS, along with high bandwidth efficiency
Cellular
FeaturesNetwork
•Hundreds of thousands of nodes
•Nodes integrate sensors, processors, transceivers withlimited resources
•Nodes are generally stationaryafter deployment
•Each node depends on small low-capacity battery as energy source, and cannot expect replacement
•The main goal is to prolong the lifetime of the network
WSN Features
9
Why WSN?
• Why Distributed Sensing?1. Dispersive media
2. Obstructions
3. Detection theory
• Why Wireless?1. Environmental lacks
10
Lifetime:Devices Power Consumption
• Sensing circuitry• Digital processing
• Radio transceiver
Communication dominates energy budgetExample: With 3J a node could transmit 1 Kb a
distance of 100 meters or efficiently execute 3 million instructions
11
Collaborative Processing (1)
• Nodes organize themselves for purposes of sensingtheir field of view, and pass information on to some users
• Goals: high detection probability, false alarm rates
Options:
1. Send raw data to a central site
2. Each node perform computation procedures to come to some decisions
12
Collaborative Processing (2)
• Architecture should limit the informationthat must flow over the network to conserve battery life and to avoid overwhelming the users
3. Signal Processing Hierarchy:
Lower false alarm probability
The load on nodes up the communicationchain is reduced
13
Collaborative Processing (3)
More reliability, Lower costHomogeneous
Elements Increased functionality: GPS or longer rangeradios need not be implemented in everyelement
Heterogeneous
Uncertain savings on hardware costsSpecial Nodes
Savings in overall network power consumption, because routing can be mademore flexible and dynamicIncreased reliability
All nodesActiveProcessors
Raw data will be tagged with timestamp and uploaded to a CN (e.g. beamforming)
Long streamsCoherent
Raw data will be preprocessed at each node toextract a set of parameters (e.g. data fusion)
Low data trafficNoncoherent
Processing
14
Collaborative Processing (4)
• Beside the design chances presented in the previous slides, two design principlesemerge from the effort to achieve reliabledecisions with low energy consumption:
1. Play the probability game to the extent youhave to
2. The processing hierarchy is closelyintertwined with networking and data storage issues
15
Self-organization (1)
• Self-organization refers to the ability of the system to achieve the necessary organizationalstructures without requiring human intervention
• Organizational structure is established to enable:1. Basic sensing2. Collaborative signal processing3. Communication network operations to support internode
and sensor system/user interaction4. Resource management
16
Self-organization (2)• Self-organization means imbue the “commander’s
intent” into the systemMany interacting devices give rise to a complex adaptive system in which Emergent Behavior isexpected
• Self-organization tasks:1. Bring the initial system online
2. Establish needed end-to-end circuits
2. Allow new nodes to be added and reconfigure when existing nodesfail
4. Quickly evolve so as to achieve these functions via low power operations
17
Novel Design Features
1. Data-centric
• Identity Data
2. Application-specific
• Intermediate nodes performapplication-specific tasks
18
Summary of Design Features
• Computation intensive, less communication
• Signal processing hierarchy
• Perform only to the extent we have to
• Processing hierarchy is intertwined with networking
• Self-organization
• Data-centric and application-specific design
19
PART 2
Solutions for networking support in Wireless Sensor Networks
20
Protocol Stack
PHY
MAC
Adaptive topology
Data dissemination
In-network aggregation
USER QUERIES
Physical
Data-link
Application
Network
MGMT
21
• Routing fidelity is maintained as long as any intermediate node is awake
• Robustness is diminished
• Needs a location information system (GPS)
r=f(R)
Adaptative Topology: GAF(2) Geographic Adaptive Fidelity
• Node equivalence is determined by dividing the area in “virtual grids”
22
Information Gathering Models
• Two are the most important gatheringmodels assumed for autonomous WSNs:– All peer nodes with a Base Station (with
constant power supply)– Only peer nodes with sinks
• By considering instead WSNs as a network in the Internet cloud, it is possible to explainthem as a set of database servers
23
Why not an Internet end to end architecture?
• Internet routes using IP address and LookupTables– Humans get data by “naming data” to a search
engine
– Many levels of indirection between data name and IP address
– Works well in Internet
• Embedded, energy-constrained, unattended, untethered systems cannot toleratecommunication overhead of indirection
24
Data Dissemination Protocols
• WSNs protocols should be:– Application specific
– Data centric
– Capable of aggregating data
– Capable of optimizing energy consumption
• The suitability of network protocolsdepends on network topology and radio parameters of the system
25
Classification of Dissemination Protocols (1)
• Based on the type of target applications and mode of operation (Agrawal, Manjeshwar) :– Proactive
– Reactive
– Hybrid
• Based on network organizational structure:– Clustered
– Flat• Hierarchical
26
Data Dissemination
Multicast
SPIN
Flooding
Unicast
Gen. Routing Protocols Proactive Reactive Hybrid
Gossiping
Clustered Clustered ClusteredFlat Flat Flat
SAR LEACH Direct
Pegasis
TEEN APTEEN‡ Directed-Diffusion*
‡ This protocol provides support for User Queries
• This protocol provides support for In-network Aggregation and User Queries
27
Proactive Clustered Protocols (1)
• Base station (far)• Localized coordination and control for cluster set-up and
operation• Local aggregation
• LEACH:– Adaptive dynamic clustering– Cluster-heads create schedule for the nodes in their
clusters TDMA– Cluster-heads perform data aggregation– Cluster-heads send data to the BS
28
• Grant a great dynamicity to the network and save the power required for organizationalproposes
• PEGASIS:– LEACH evolution– Only one leader at a turn– Communication chain
• Advantages:– Shorter distances – Only one leader
Proactive Flat Protocols
c1
c0
c4
c3
c2
BS
1
2
3
46
5
7
29
Reactive Clustered Protocols
• Only when sensed data exceeds a threshold value, an alerting message is sent to the interested nodes
• Pay particular attention to time critical attributes
• Generally these protocols outperform proactive onesunder an energy dissipation viewpoint
• TEEN:– LEACH based initialization
– Hard and soft thresholds
30
Multicast Data Dissemination• Flooding
1. Implosion
2. Overlap
3. Resource blindness
• Gossiping2. Avoids implosion problem
(1) (2)
31
• SPIN:
Multicast Data Dissemination
1. Negotiation 2. Meta-data 3. Resource Adaptation
32
In-network aggregation• The goal is to facilitate the communication among
sources and sinks
• Directed Diffusion:– Not host based but data-centric Application-specific
attribute based naming
– In-network processing through application specificfilters
– Localized interactions
– Trade-off:
Energy efficiency vs. Robustness and Scalability– Data rate proactive
– Attribute-value pairs event-driven
33
Localized Algorithms (1)
• Collaborative and distributed computation in which sensor nodes communicate with sensorswithin some neighborhood, yet the overallcomputation achieves a desired globalobjective
• Properties:– Scalability
– Robustness
34
Localized Algorithms (2)
• Design is hard:– Global behavior
– Parametrical Sensitivity
• Approaches to overcome these difficulties:
– Develop intuition by prototyping
– Develop techniques for characterizing the
performance
35
Application Example: Remote Surveillance
• Interrogation:
–– e.g., “Give me periodic reporte.g., “Give me periodic reportss about animal about animal location in region A every t seconds”location in region A every t seconds”
• Interrogation is propagated to sensor nodes in
region A
• Sensor nodes in region A are tasked to collect data
• Data are sent back to the users every t seconds
Basic Directed Diffusion
Source
Sink
Interest = Interrogation
Gradient = Who is interested
CLASS_KEY IS INTEREST_CLASSLONGITUDE_KEY GE 10LONGITUDE_KEY LE 50LATITUDE_KEY GE 100LATITUDE_KEY LE 120SENSOR EQ MOVEMENTINTENSITY GE 0.6CONFIDENCE GE 0.7INTERVAL IS 10EXPIRE_TIME IS 100
Basic Directed Diffusion
Source
Sink
Interest = Interrogation
Gradient = Who is interested
2. subscribe (AttrVec, ApplCallback)1. subscribe (InterestAttrVec, Callback)
InterestAttrVecCLASS_KEY EQ INTEREST_CLASSLONGITUDE_KEY IS 35LATITUDE_KEY IS 110SENSOR IS MOVEMENT
3. addFilter (FilAttrVec, FilterCallback)
FilterAttrVecCLASS_KEY EQ DATA_CLASSSENSOR EQ MOVEMENTINTENSITY GE 0.7
Basic Directed DiffusionInterests Setting up gradients
Source
Sink
Interest = Interrogation
Gradient = Who is interested
39
Basic Directed Diffusion
Source
Sink
4. h = publish (SensedAttrVec)5. send (h, SensedAttrVec)
SensedAttrVecCLASS_KEY IS DATA_CLASSLONGITUDE_KEY IS 35LATITUDE_KEY IS 110SENSOR IS MOVEMENTINTENSITY IS 0.8CONFIDENCE IS 0.7
Low rate event
Sending data …
40
Basic Directed Diffusion
Source
Low rate event
6. FilterCallback.recv (Message m1)
m2CLASS_KEY IS DATA_CLASSLONGITUDE_KEY IS 35LATITUDE_KEY IS 110SENSOR IS MOVEMENTINTENSITY IS 0.8CONFIDENCE IS 0.8
7. sendMessage (Message new)
m1a
m1b
m2
m2
41
Basic Directed Diffusion
Source
Sink
Low rate event
8. ApplCallback.recv (NRAttrVec)
Basic Directed Diffusion
Source
Sink
… and Reinforcing the best path
Low rate event Reinforcement = Increased interest
CLASS_KEY IS INTEREST_CLASSLONGITUDE_KEY GE 10LONGITUDE_KEY LE 50LATITUDE_KEY GE 100LATITUDE_KEY LE 120SENSOR EQ MOVEMENTINTENSITY GE 0.6CONFIDENCE GE 0.7INTERVAL IS 1EXPIRE_TIME IS 90
Directed Diffusion and Dynamics
Recoveringfrom node failure
Source
Sink
Low rate event
High rate eventReinforcement
Directed Diffusion and Dynamics
Source
Sink
Stable path
Low rate event
High rate event
Directed Diffusion and Dynamics
Recoveringfrom link failure
Source
Sink
Low rate event
High rate eventReinforcement
Directed Diffusion and Dynamics
Stable path
Source
Sink
Low rate event
High rate eventReinforcement
Use: “Interests set up gradients drawing down data”
47
Query Models
• Historical Queries: they are mainly used for analysisof historical data– E.g. “What was the watermark two hours ago in the
southeast?”
• One-time Queries: they give a snapshot view of the network– E.g. “What is the watermark in the southeast?”
• Persistent Queries: they are used for monitoringtasks over a time interval– E.g. “Report the watermark in the southeast for the next
four hours”
User QueriesDatabase Oriented Approaches
• Assumptions:– WSNs have the capability to forward packets in an
autonomous manner
– Each node runs a mini-server
Warehouse
Front-End
Sensor Nodes
Traditional CentralizedApproach
Sensor
DB
Front-End
Sensor Nodes
SenDB
SenDB
SenDB
SenDB
SenDB
Sensor DatabaseSystem
49
• COUGAR:– Devices ADTs e.g. RFSensor(Sensor, X, Y)– SQL semantic extended to include new query
types
– New possible plans (location)
– New metrics (resource usage and reaction time)
• Other issues:– Stream processing
– Quality of Service (latency vs. completeness)
– In-network processing and aggregation
User QueriesDatabase Oriented Approaches
50
Conclusions and On-going Work
• Lack of capabilities of devices– New design guidelines– New models (protocol stack)
• On-going work– Energy harvesting techniques– Enlarge testbeds– Develop new applications
• Mobile code-based management• Interactions with Internet
51
References
• WSN Concepts– I. Akyildiz, W. Su, Y. Sankarasubramaniam, E. Cayirci, “A Survey
on Sensor Networks”, IEEE Communications Magazine, Aug. 2002, pp. 102-114
– J. Pottie, “Wireless Sensor Networks”, ITW 1998, pp. 139-140– J. Pottie, W. Kaiser, “Wireless Integrated Network Sensors”,
Communications of the ACM, May 2000, pp. 51-58– J. Pottie, “Hierarchical Information Processing in Distributed
Sensor Networks”, ISIT 1998, p. 163
• Self-organization– L. Clare, J. Pottie, J. Agre, “Self-Organizing Distributed Sensor
Networks”, SPIE Conf. Unattended Ground Sensor Technologies and Applications 1999, pp. 229-237
52
References
• Solutions for networking support• SMACS
– K. Sohrabi, J. Gao, V. Ailawadhi, J. Pottie, “Protocols for Self-Organization of a Wireless Sensor Network”, IEEE Personal Communications, Oct. 2000, pp. 16-27
• GAF– Y. Xu, J. Heidemann, D. Estrin, “Geography-Informed Enery
Conservation for Ad Hoc Routing”, MOBICOM 2001, pp. 70-84
• LEACH– W. Heinzelman, A. Chandrakasan, H. Balakrishnan, “Energy-
efficient Communication Protocols for Wireless MicrosensorNetworks”, Hawaiian Int. Conf. Systems Science 2000
53
References
• Pegasis– S. Lindsey, C. Taghavendra, K. Sivalingam, “Data Gathering Algorithms
in Sensor Networks Using Energy Metrics”, IEEE Transactions on Parallel and Distributed Systems, Sep. 2002, pp. 924-935
• TEEN– A. Manjeshwar, D. Agrawal, “TEEN: A Routing Protocol for Enhanced
Efficiency in Wireless Sensor Networks”, Int. Workshop Parallel and Distributed Computing Issues in Wireless Networks and Mobile Cokmputing 2001
• APTEEN– A. Manheshwar, D. Agrawal, “An Analytical Model for Information
Retrieval in Wireless Sensor Networks Using Enhanced APTEEN Protocol”, IEEE Transaction on Parallel and Distributed Systems, Dec. 2002, pp. 1290-1302
54
References• SPIN
– W. Rabiner Heinzelman, J. Kulik, H. Balakrishnan, “Adaptive Protocolsfor Information Dissemination in Wireless Sensor Networks”, MOBICOM 1999, pp. 174-185
• Localized Algorithms– D. Estrin, R. Govindan, J. Heidemann, S. Kumar, “Next Century
Challenges: Scalable Coordination in Sensor Networks”, MOBICOM 1999, pp. 263-270
• Directed Diffusion– J. Heidemann, F. Silva, C. Intanagonwiwat, R. Govindan, D. Estrin, D.
Ganesan, “Building Efficient Wireless Sensor Networks with Low-LevelNaming”, ACM Symp. Operating Systems Principles 2001, pp- 146-159
– C. Intanagonwiwat, R. Govindan, D. Estrin, J. Heidemann, F. Silva, “Directed Diffusion for Wireless Sensor Networking”, IEEE/ACM Transactions on Networking, Feb. 2003, pp. 2-16
• COUGAR– P. Bonnet, J. Gehrke, P. Seshadri, “Towards Sensor Database Systems”,
Int. Conf. Mobile Data Management 2001
55