Adaptive Protocols for Information Dissemination in Wireless Sensor NetworksOutlineIntroductionIntroduction contd.Sensor Protocols for Information via Negotiation (SPIN)Implosion ProblemOverlap problemSPIN contd..SPIN Meta-DataSPIN MessagesSPIN-1 and SPIN-2Node B sends a REQ listing all of the data it would like to acquire.If node B had its own data, it could aggregate this with the data of node A and advertise.Nodes need not respond to every messageSPIN-2Other data dissemination algos.PowerPoint PresentationIdeal disseminationSlide 19Sensor Network SimulationsSimulation TestbedSPIN-1 ResultsData Acquired Over TimeEnergy Dissipated Over TimeSlide 25Unlimited Energy SimulationsLimited Energy SimulationsLimited Energy Simulations contd..Best-Case Convergence TimesTransmission time per data packet = 8s/d Since SPIN-1 has to process ADV, REQ, DATA so processing time = 3(d+r)Convergence Time – overlapping dataSlide 32ConclusionsStrengths and WeaknessesQuestions ?CS 599 Intelligent Embedded Systems 1Adaptive Protocols for Information Dissemination in Wireless Sensor NetworksW.R.Heinzelman, J.kulik, H.BalakrishnanCS 599 Intelligent Embedded Systems 2OutlineIntroductionSPINOther Data Dissemination AlgorithmsSensor Network SimulationsConclusionsStrengths and WeaknessesCS 599 Intelligent Embedded Systems 3IntroductionWide deployment of Wireless sensor networksWireless sensor networksCan aggregate sensor data to provide multi-dimensional viewImprove sensing accuracyFocus on critical events (e.g. intruder entering)Fault tolerant networkCan improve remote access to sensor data – sink nodesCS 599 Intelligent Embedded Systems 4Introduction contd.Limitations of Wireless sensor networksEnergyComputationCommunicationCS 599 Intelligent Embedded Systems 5Sensor Protocols for Information via Negotiation (SPIN)Classic flooding limitationsImplosionOverlapResource blindnessCS 599 Intelligent Embedded Systems 6Implosion ProblemCS 599 Intelligent Embedded Systems 7Overlap problemCS 599 Intelligent Embedded Systems 8SPIN contd..SPIN overcomes these deficienciesNegotiationResource-adaptationEach sensor node has resource managerKeeps track of resource consumptionApplications probe the manager before any activityCut down activity to save energyMotivated by principle of ALFCommon data naming (meta-data)CS 599 Intelligent Embedded Systems 9SPIN Meta-DataSensors use meta-data to describe the sensor data brieflyIf x is the meta-data descriptor for data Xsizeof (x) < sizeof (X)If x==ysensor-data-of (x) = sensor-data-of (y)If X==Ymeta-data-of (X) = meta-data-of (Y)Meta-data format is application specificCS 599 Intelligent Embedded Systems 10SPIN MessagesADV – new data advertisementREQ – request for dataDATA – data messageADV and REQ messages contain only meta-data so they are smaller in size.CS 599 Intelligent Embedded Systems 11SPIN-1 and SPIN-2SPIN-1Simple 3-stage handshake protocolData aggregation is possibleCan adapt to work in lossy or mobile networkCan run in a completely unconfigured networkCS 599 Intelligent Embedded Systems 12Node B sends a REQ listing all of the data it would like to acquire.CS 599 Intelligent Embedded Systems 13If node B had its own data, it could aggregate this with the data of node A and advertise.CS 599 Intelligent Embedded Systems 14Nodes need not respond to every messageCS 599 Intelligent Embedded Systems 15SPIN-2SPIN-1 with a Low-Energy ThresholdWhen energy below energy threshold – stop participating in the protocolCan just receive data avoiding ADV-REQ phaseCS 599 Intelligent Embedded Systems 16Other data dissemination algos.Classic FloodingConverges in O(d), d-diameter of the networkGossipingForward data to a random neighborAvoids implosionDisseminates at a slow rateFastest rate = 1 node/roundCS 599 Intelligent Embedded Systems 17CS 599 Intelligent Embedded Systems 18Ideal disseminationEvery node sends sensor data along shortest pathReceives each piece of distinct data only onceImplementationNetwork level multicast (source specific)To handle losses, reliable multicast has to be deployedSPIN is a form of application-level multicastCS 599 Intelligent Embedded Systems 19CS 599 Intelligent Embedded Systems 20Sensor Network SimulationsSimulated using ns simulatorExtended ns to create a Resource-Adaptive NodeCS 599 Intelligent Embedded Systems 21Simulation TestbedCS 599 Intelligent Embedded Systems 22SPIN-1 ResultsHigher throughput than gossipingSame throughput as floodingUses substantially less energy than other protocolsSPIN-2 delivers more data per unit energy than SPIN-1SPIN-2 performs closer to Ideal disseminationNodes with higher degree tend to dissipate more energy than nodes with lower degreeCS 599 Intelligent Embedded Systems 23Data Acquired Over TimeCS 599 Intelligent Embedded Systems 24Energy Dissipated Over TimeCS 599 Intelligent Embedded Systems 25Energy Dissipated Over TimeCS 599 Intelligent Embedded Systems 26Unlimited Energy SimulationsCS 599 Intelligent Embedded Systems 27Limited Energy SimulationsCS 599 Intelligent Embedded Systems 28Limited Energy Simulations contd..CS 599 Intelligent Embedded Systems 29Best-Case Convergence TimesFor overlapping sensor dataConvergence times for ideal and flooding are the sameFor non-overlapping sensor dataFlooding converges faster than SPIN-1To understand these results, we develop equations that predict convergence times of each of these protocols.CS 599 Intelligent Embedded Systems 30Transmission time per data packet = 8s/dSince SPIN-1 has to process ADV, REQ, DATA so processing time = 3(d+r))8)(3()83()8(,)8(1bsrdlCbsdlbsrdlCCbsdldSPINddFloodIdealdConvergence Time – no overlapCS 599 Intelligent Embedded Systems 31Convergence Time – overlapping data24)1()8()83()8)(3()83()8(,)8(1rbskdbskrdlbsdlbskrdlCbsdlbskrdlCCbsdllplplpSPINlplpFloodIdeallpCS 599 Intelligent Embedded Systems 32For the testbed network parametersSimulation resultsFlooding converges in 135msIdeal converges in 125msSPIN-1 converges in 215msConvergence times of flooding and ideal are closer to their upper bound unlike SPIN-1294.0133.0154.0,063.01SPINFloodIdealCCCCS 599
View Full Document