2 Wireless Sensor and Actor Networks I. F. Akyildiz and I. H Wireless Sensor and Actor Networks I.F. Akyildiz and I. H. Kasimoglu,“Wireless Sensor and Actor Networks: Research Challenges” Ad Hoc Networks Journal (Elsevier), pp , OctTask ManagerNodeSinkSensor/Actor FieldSensorsActors
3 Actuators vs. Actors Why do we call them actors? Actuator (Texas Instruments Technical Glossary):“An actuator is a device to convert an electrical control signal to a physical action. Actuators may be used for flow-control valves, pumps, positioning drives, motors, switches, relays and meters.”The mobility of a robot may be enabled by several actuators (motors, servo-mechanisms, etc)However, the robot represents one single network entity which we refer to as actorHence, one actor can be endowed with multiple actuators
4 Wireless Sensor and Actor Networks SensorsPassive elements sensing from the environmentLimited energy, processing and communication capabilitiesActorsActive elements acting on the environmentHigher processing and communication capabilitiesLess constrained energy resources (Longer battery life or constant power source)
5 Sub-Kilogram Intelligent Tele-robots (SKITs): Networked Robots having Coordination & Wireless Communication Capabilities
6 Robotic Mule: Autonomous Battlefield Robot designed for the Army
7 Mini-Robot(developed at Sandia National Laboratories)
9 Components of Sensor & Actor Nodes SensingUnitProcessor& StorageTransceiverADCSensor NodePower UnitActuationUnitController(DecisionUnit)Processor& StorageTransceiverDACActor NodePower Unit
10 Integrated Sensor & Actor Nodes ProcessingSensingUnitActuationControllerDecision ProcessPower UnitTransceiver
11 WSAN Applications Environmental Applications: Detecting and extinguishing forest fire.Microclimate control in buildings:In case of very high or low temperature values, trigger the audio alarm actors in that area.Distributed Robotics & Sensor Network:(Mobile) robots dispersed throughout a sensor network alarm actors in that area.
12 WSAN Applications Parking Airport Safety City Maintenance Sewage and Contamination ControlBattlefield Applications:Sensors detect mines or explosive substancesActors annihilate them or function as tanks
13 WSANs vs. Wireless Sensor Networks Real-Time Requirements for Timely ActionsRapidly respond to sensor input (e.g., fire application)To perform right actions, sensor data must be valid at the time of actingHeterogeneous Node DeploymentSensorsActorsDensely deployedLoosely deployed due to the differentcoverage requirements and physicalinteraction methods of acting task
14 WSANs vs. Wireless Sensor Networks Coordination RequirementsSensor-Actor CoordinationActor-Actor Coordination
15 WSAN Communication Architecture Semi-Automated Architecture SinkSensors Sink ActorsRequires manual intervention at sinkNo sensor-actor and actor-actor coordination neededSimilar to the conventional WSN architectureEvent Area
16 WSAN Communication Architecture Automated Architecture SinkEvent AreaSensors ActorsNo intervention from sink is necessaryLocalized information exchangeLow latencyDistributed sensor-actor and actor-actor coordination required
17 SENSOR-ACTOR COORDINATION Challenges:Which sensor(s) communicate with which actor(s) (Single or Multiple Actors)How should the communication occur? (i.e., single-hop or multi-hop)What are the requirements of the communication (i.e., real-time, energy efficiency)
18 Sensor-Actor Coordination Which sensor(s) communicate with which actor(s)?CASE 1:Minimum number of sensors to report the sensed eventCASE 2:Minimum set of actors to cover the event regionBoth cases aboveThe entire set of sensors and actors in the vicinity of the regionThe set of actors whose acting regions do not overlap
19 Sensor-Actor Coordination SINGLE ACTOR Event AreaSelection of the most appropriate actorTo select, sensors need to coordinate with each other
20 Sensor-Actor Coordination SINGLE ACTOR Selecting a single actor node may be based on:The distance between the event area and the actorThe energy consumption of the path from sensors to the actorThe action range of the actor
21 Sensor-Actor Coordination MULTI ACTORS Event AreaActorClustering is requiredSensors only need to coordinate with sensors within some neighborhood to form clusters or groups
22 Sensor-Actor Coordination MULTI ACTORS Clusters may be formed such a way thatThe event transmission time from sensors to actors is minimizedThe events from sensors to actors are transmitted through the minimum energy pathsThe action regions can cover the entire event area
23 ACTOR-ACTOR COORDINATION Challenges:Which actor(s) should execute which action(s)?How should multi-actor task allocation be done?
24 Actor-Actor Coordination Single-Actor Task vs. Multi-Actor TaskSingle-Actor TaskHow is the single actor selected?Multi-Actor TaskWhat is the optimum number of actors performing actions?Selection of most fit actors among the capable actors for that taskOnly a subset of actors covering the entire event region may perform the task to save action energy
25 A Distributed Coordination Framework for WSANs T. Melodia, D A Distributed Coordination Framework for WSANs T. Melodia, D. Pompili, V. C. Gungor, I. F. Akyildiz, ACM MOBIHOC’05, May Also in IEEE Transactions on Mobile Computing, 2007.Comprehensive framework for coordination problemsSENSOR-ACTOR COORDINATIONOptimal Event-driven ClusteringA Distributed Scalable ProtocolACTOR-ACTOR COORDINATIONOptimal SolutionReal-time Localized Auction
26 Coordination Requirements Sensor-Actor CoordinationEstablish data paths between sensors and actorsMeet energy efficiency and real time requirementsActor-Actor CoordinationDecision: Does an action need to be performed?Which action should be performed?How to share the workload among actors?
27 Sensor-Actor Coordination Objectives:Establish data paths between sensors and actorsMeet energy efficiency and real-time requirementsQuestion:To which actor does each sensor send its data?Solution:Event Driven Clustering with Multiple Actors
28 Event-Driven Clustering with Multiple Actors Event Area1. Event Occurs2. Sensor-Actor Coordination: Event-Driven ClusteringWhat is the optimal clustering strategy? Distributed algorithm?
29 Reliability Definition 1. The latency bound B is the maximum allowed time between sampling of the physical features of the event and the moment when the actor receives a data packet describing these event features
30 Reliability Definition 2 A data packet is EXPIRED (UNRELIABLE), if it does not meet the latency bound BDefinition 3A data packet is UNEXPIRED (RELIABLE), if it is received within the latency bound B
31 Reliability Definition 4: The event reliability r is the ratio of reliable data packets over all packets received in a decision intervalDefinition 5:The event reliability threshold rth is the minimum event reliability required by the applicationOBJECTIVE:Comply with the event reliability threshold (r>rth) with minimum energy expenditure!
32 Event-Driven Clustering with Multiple Actors Objective:Find the optimal strategy for event-driven clustering (To which actors is data sent? Which paths are used?) a joint Clustering and Routing problem
33 Event-Driven Clustering with Multiple Actors Requirements of the Optimal Solution:Provide reliability above the event reliability threshold (r>rth)Minimize overall Energy ConsumptionOptimal solution obtained by Integer Linear Programming formulation
34 Event-Driven Clustering with Multiple Actors ILP Formulation is provided -> allows finding the optimal solutionBUT NP-Complete problem:Not scalable (<100 nodes)Centralized solutionHelps gaining insight in the properties of the optimal solutionPerformance benchmark for distributed, more scalable solutions
35 A Distributed Protocol Find the optimal working point of the network, i.e.:r>rth ( reliability over the threshold)Minimum energy consumptionBased on the feedbacks from actors:Actor calculates reliability r and broadcasts its value to the sensors
36 A Distributed Protocol If the reliability r is complied with (r>rth), a certain portion of the sensors switch in the aggregation state to save energy (lower energy consumption, higher delay)Equilibrium is reached when reliability threshold is met (r ≈ rth) with minimum energy consumption.
37 A Distributed Protocol BASIC IDEA:When the event is first sensed, sensors all begin in the start-up state and establish data paths to the actorsIf reliability is advertised to be low (r<rth)Certain portion of the sensors switch to speed-up state, which shortens the end-to-end paths (lower delay, higher energy consumption)
38 A Distributed Protocol Sensors probabilistically switch among three different states according to feedback from the actors:Start-up State:Quickly establish a data path from each source to one actorCompromise between energy consumption and latency
39 A Distributed Protocol Speed-up State:Reduce the number of hops in sensor-actor paths so as to reduce the end-to-end delayObtained by sending packets to “far” neighbors (closer to the destination actor)
40 A Distributed Protocol Aggregation state:Reduce the overall energy consumption when compliant with event reliabilitySend packets to closer neighbors (higher number of hops)
41 Example: Path Establishment nodes establish paths (start-up state)idle start-up statean event occursAnother actor is too far away and thus not energy efficient for any of the nodes in the event area
42 Example: Low Reliability Some sensors switch to the speed-up state (probabilistically) and select as next hop the closest node to the actor reduce latencyThe actor advertises low reliability (r<rth)idle start-up state speed-up state
43 Example: High Reliability Some sensors switch to the aggregation state (probabilistically) and select as next hop the closest node already in the da-tree reduce energy consumptionThe actor advertises high reliability (r>rth)idle start-up state speed-up state aggregation state
44 Actor-Actor Coordination Objective:Selecting the best actor(s) in terms of action completion time and energy consumption so as to perform the action!Challenges:Which actor(s) should execute which action(s)?How should multi-actor task allocation be done?
45 Actor-Actor Coordination Model DEFINITIONs:Overlapping Area:Area can be acted upon by multiple actorsNon-Overlapping Area:Area can be acted upon only by one actor
46 Actor-Actor Coordination Model Action Completion Time Bound:The maximum allowed time from the moment when the event is sensed to the moment when the action is completedPower Levels:Discrete levels of power for performing the action A higher power level corresponds to a lower action completion time!
47 Actor-Actor Coordination Problems For an Overlapping Area, actor-actor coordination problem:Selecting a subset of actorsAdjusting action power levels Maximize the residual energy and complete the action within the action completion bound
48 Actor-Actor Coordination Problems For a Non-Overlapping Area, actor-actor coordination problem:Adjusting action power levels Maximize the residual energy
49 Actor-Actor Coordination Optimal Solution:Actor-actor coordination problem formulated as a Residual Energy Maximization Problem using Mixed Integer Non-Linear Programming (MINLP)Distributed Solution:Real-Time Localized Auction-Based Mechanism
50 Real-Time Localized Auction-Based Mechanism Inspired by the behaviors of agents in a Market Economy Interactions between buyers and sellersPossible Roles of the Actors:Seller: Actor receiving the event featuresAuctioneer: Actor in charge of conducting the auctionBuyer: Actor(s) that can act on a particular overlapping area
51 Real-Time Localized Auction-Based Mechanism For overlapping areas:Seller selects one auctioneer for each overlapping area, i.e., the closest actor to the center of the overlapping area Energy spent for auction and auction time reduced!Seller informs each auctioneer about the auction area and the action time bound
52 Real-Time Localized Auction-Based Mechanism Auctioneer determines the winners of the auction based on the bids received from the buyers.Bids consists of available energy, power level and action completion time
53 Real-Time Localized Auction-Based Mechanism The auctioneer finds the winners by calculating the optimal solution of the Residual Energy Maximization ProblemFor Non-Overlapping areasThe corresponding actor is directly assigned the action task
54 Sensor-Actor Coordination Start-up (speed-up) configuration: all nodes are in the start-up (speed-up) stateComparison between the optimal solution of the event-driven clustering problem and the energy consumption of start-up, speed-up, aggregation configuration with varying event ranges (60 sensors; 4 actors)
55 Sensor-Actor Coordination Comparison of the energy consumption of different configurationsThe energy consumption in the aggregation configuration is much lower that in the start-up and speed-up configuration
56 Sensor-Actor Coordination Comparison of average number of hops for start-up and speed-up configuration. The speed-up configuration shows paths with lower delay (less hops)
57 Cyber Physical Systems Integration of computation with physical processes. Embedded computers and networks monitor and control physical processes in feedback loops where physical processes affect computations and vice versa. CPS will blend sensing, actuation, computation, networking, and physical processes as action networks."Networked Information Technology Systems Connected With The Physical World", also referred to as cyber-physical systems, are cited as the top technical priority for networking and IT research and development.President's Council of Advisors on Science and Technology