Adaptive Protocols for Information Dissemination in Wireless Sensor Networks

1. Adaptive Protocols for Information Dissemination in Wireless Sensor Networks Authors: Wendi Heinzelman, Joanna Kulik, and HariBalakrishnanMassachusetts Institute of Technology

2. Overview • Introduction • SPIN Protocols • Simulation results • Conclusion

3. Introduction • Why have wireless sensor networks? • Extend ability to monitor & control physical environment from remote locations • Improve the sensing accuracy through distributed processing of vast information via collaboration among sensor nodes and online information processing at those nodes.

4. Introduction… • Why have wireless sensor networks? …. • Provide multi-dimensional view of environment • Focus attention on critical events pointed out by other sensors in the network • Networked sensors can continue to function accurately even with the failure of individual sensors

5. Introduction… • Some differences between adhoc and sensor networks. • Number of nodes can be several orders of magnitude higher • Sensor nodes are densely deployed • Sensor nodes are prone to failure • Sensor nodes are limited in power, computing capacity and memory • Sensor nodes may not have global identification • Sensor nodes are designed for specific applications • Each node is capable of collecting data and routing data back to the sink • Sink may communicate via Internet/Satellite with the task manager

6. Design factors • Fault-tolerance • Nodes may fail or be blocked due to lack of power, or environmental interference • Failure of some nodes should not affect the overall task of the network • Scalability • Depends on the application, sensor nodes deployed may run in thousands or millions • Density of nodes can range from few to few hundreds in a region, which can be less than 10m in diameter • Production costs • Cost of a single node is important for deployment

7. Design factors… • Hardware Constraints • A sensor node is made up of four basic components • Sensing unit: consists of a sensors and analogue to digital converters(ADC). ADCs convert the analogue signals produced by the sensors to digital signals and then fed into the processing unit. • Processing unit: manages the procedures that make the sensor node collaborate with other nodes to carry out assigned tasks. • Transceiver unit: connects the node to the network • Power unit: may be solar cells or batteries • It is also common that the node has a location finding unit such as GPS • All the units should fit into a matchbox size module.

8. Design factors… • Issues related to topology maintenance • Predeployment and deployment phase: sensor nodes can be thrown as a mass or placed one by one • Post deployment phase: After deployment, topology changes are due to energy, malfunction, failure, … • Redeployment of additional nodes: Additional sensor nodes can be redeployed at anytime to replace malfunctioning nodes or due to changes in the task dynamics.

9. Design factors… • Environment: Sensor nodes are placed either very close to or inside the phenomenon to be observed • May be interior of large machinery, at the bottom of the ocean, in a chemically contaminated field, battlefield, …. • Transmission media: Wireless links such as infrared or radio • Power consumption: replacement of power sources may be impossible. • A node will have to serve as both data generator and router • So, we need power-aware protocols such as power-aware routing

10. Protocol stack • Physical layer: responsible for frequency selection, signal detection, etc • Energy efficient solutions at this layer are essential • Data link layer: responsible for multiplexing of data stream, media access control, and error control. New Mac protocols that are suitable for sensor networks need to be designed • Researchers have already proposed some protocols in this direction

11. Protocol stack… • Network Layer: Some design principles for this layer are • Power efficiency is an important consideration • Sensor networks are data centric • Data aggregation is useful only if it does not hinder the collaborative effort of the sensor nodes • An ideal sensor network has attribute based addressing and location awareness

12. Network Layer… • Design principles for designing efficient routing algorithms • Power available (PA) on the route • Energy required for transmission using the route • Criteria for selecting a route: • Select a route with maximum available power • Select a route that uses minimum energy • Select a route along which the minimum PA is larger than the minimum PA of other routes chosen. This prevents the risk of using up the power at a sensor node with low PA much earlier than others because they are on a route with nodes that have high PA.

13. Design of Routing protocols… • Data-centric approach for routing: • Interest dissemination is performed to assign sensing tasks to the sensor nodes • Two approaches for interest dissemination: • Sink nodes broadcast interest • Sensor nodes advertise for available data and wait for interest from available nodes • Data-centric routing algorithm requires attribute-based naming • For example, “areas where temperature is > 70 degrees” • Data aggregation(fusion), a technique used for solving implosion and overlapping problem of data-centric routing • Data coming from multiple nodes is aggregated at the intermediate nodes on the way to sink.

14. Current research in the network layer • Flooding: each node receiving the data or management packet is rebroadcasts it until it reaches the destination or travels maximum number of hops • Problems with this approach: • Implosion: duplicate messages are sent by the same node • Overlap: If two nodes share the same observing region, both of them may sense the same stimuli at the same time • Resource blindness: does not take into account available energy resources

15. Network layer…. • Gossiping: a node randomly selects one of its neighbors to send data, which in turns chooses one of its neighbor randomly • Minimizes the implosion problem, but takes long time to propagate the message to all nodes

16. In the Future… • Collection of sensor nodes form ad hoc distributed processing networks about the physical environment • Each sensor node operates autonomously with no central point of control in the network • Each node bases its decision on … • its mission • info it currently possesses • knowledge of computing, communication & energy resources • Today is isolated sensors – tomorrow’s networked sensors have more potential for performance with more accuracy, robustness, and sophistication

17. What is SPIN? • SPIN stands for Sensor Protocols for Information via Negotiation • Family of adaptive protocols that efficiently disseminate information among sensors in an energy-constrained wireless sensor network.

18. SPIN is… • Family of negotiation-based information dissemination protocols suitable for wireless sensor networks. • Focus on efficient dissemination of individual sensor observations to all the sensors in the network • Treat all sensors as potential sink nodes.

19. Benefits of solving this problem • Gives us a way to replicate complete views of the environment across the network to enhance fault-tolerance of the system • A way of disseminating a critical piece of information to all nodes (such as an intrusion has been detected in a surveillance network)

20. SPIN… • SPIN tries to address the problem caused by protocols that use classic flooding. As we know such protocols have the following problems: • Implosion • Overlap • Resource blindness

21. SPIN Solution Spin overcomes classic flooding problems by • Negotiation • Deals with implosion and overlap problems by having nodes negotiate with each other before transmitting data • Utilizes data descriptors to identify data called meta-data which is used during negotiations • Resource-adaptation • Addresses the resource blindness problem by having nodes poll their resources before data transmission. • Resource Manager for each node

22. Motivation for the design of the protocols • Application Level Framing also known as ALF • Network protocols must choose transmission units that are meaningful to applications. • Packetization is best done in terms of Application Data Units (ADUs) • Important component of ALF-based protocols is common data naming between the transmission protocol & application which the authors follow in the design of meta-data. • Take the ALF-like idea one step further by arguing that routing decisions are also best made in application-controlled and application-specific ways, using knowledge of not just the network topology but application data layout & the state of resources at each node.

23. SPIN Approach Components of SPIN • Meta-Data • SPIN Messages • SPIN Resource Management Protocols • SPIN-1: A 3-Stage Handshake Protocol • SPIN-2: SPIN-1 with a Low-Energy Threshold

24. Meta-Data • Sensors use meta-data to succinctly & completely describe the data that they collect. • Meta-data must be shorter than sensor data. • If two pieces of actual sensor data are distinguishable, then their corresponding meta-data should be distinguishable as well. • SPIN does not specify a format for meta-data. • Format is application-specific. • SPIN relies on each application to interpret and synthesize its own meta-data.

25. Meta-Data… • Drawbacks: Costs associated with storage, retrieval, and general management of meta-data. • But benefits of succinct representation for large data messages in SPIN far outweighs these costs.

26. SPIN Messages • ADV – new data advertisement • REQ – request for data • DATA – data message

27. SPIN Resource Management • SPIN applications are resource-aware and resource-adaptive. • SPIN nodes can poll their system resource to determine energy cost to make informed decisions. • SPIN does not specify a particular energy management policy • It specifies an interface that applications can use to probe their available resources.

28. SPIN-1…A 3-Stage Handshake Protocol • SPIN-1 is a simple handshake protocol for disseminating data through a lossless network. • Works in 3-stages: • ADV (New Data Advertisement) • REQ (Request for Data) • DATA (Data Message)

29. SPIN-1… • ADV Stage (New Data Advertisement) • It starts when a node obtains new data that it wants to disseminate by sending ADV message to its neighbors • REQ Stage (Request for Data) • Node sends REQ message when it wishes to receive actual data • DATA Stage (Data Message) • Initiator node responds to the sent REQ message by sending DATA to the requesting node.

30. SPIN-1… Things to keep in mind…. • If requesting node has its own data, it can aggregate this data with the data it receives & then send advertisements of this aggregated data to its neighbors • Nodes are not required to respond to every message

31. ADV SPIN-1 Protocol A B Node A starts by advertising its data to node B.

32. REQ SPIN-1 Protocol A B Node B responds by sending a request to node A.

33. DATA SPIN-1 Protocol A B Node A send requested Data to Node B.

34. ADV ADV ADV ADV ADV SPIN-1 Protocol A B After Node B receives the Data, it sends out advertisements to its neighbors.

35. REQ REQ REQ REQ SPIN-1 Protocol A B Then in return these neighboring nodes will send requests back to B.

36. DATA DATA DATA DATA SPIN-1 Protocol A B Node B sends Data to all requesting neighbors.

37. Properties of SPIN-1 Protocol • Simple • Each node only needs to know about its single-hop neighbors • No topology info required • Nodes do not have to respond to every message received • Nodes receiving data could aggregate its own data with data received from neighbors • Then advertise aggregated data to its neighbors

38. SPIN-2: SPIN-1 with a Low-Energy Threshold • Adds simple energy-conservation heuristic to the SPIN-1 protocol Cases: • High Energy - use the same 3-stage protocol as SPIN-1 • Low Energy – adapt by reducing its participation in the protocol

39. SPIN-2: SPIN-1 with a Low-Energy Threshold • A node will only participate in a stage of the protocol if it believes that it can complete all of the other stages without going below the low-energy threshold. • This approach does not prevent nodes from receiving, but only prevent nodes from ever handling DATA message below low-energy threshold.

40. Data Dissemination Algorithms Chosen for Comparison • Classic Flooding • Gossiping – Instead of indiscriminately forwarding data to all its neighbors, a gossip node forwards data to only a randomly selected neighbor. • Ideal Dissemination • Every node sends observed data along a shortest path • Every node receives each piece of distinct data only once • Data is transmitted in the shortest possible amount of time

41. (a) (a,c) 1 2 (c) (a) Ideal Dissemination (a,c) A This illustration shows ideal dissemination of observed data a & c. B C (c) D

42. Ideal Dissemination Approaches using shortest path in current networks ● Network-level multicast - IP multicast ● Reliable multicast Problems with these approaches: ● Complicated protocol machinery ● Most existing approaches to shortest- path distribution trees would have to be modified to achieve ideal dissemination; So they compare SPIN with an ideal dissemination protocol.

43. Testing SPIN Efficiency • Simulation-based study for comparison • Five dissemination protocols evaluated • SPIN-1 • SPIN-2 • Flooding • Gossiping • Ideal

44. Sensor Network Simulations • ns Implementation • Simulation Testbed • Unlimited Energy Simulations • Limited Energy Simulations • Best-Case Convergence Times

45. Simulation Testbed • 25-node network • Randomly generated • Fully connected • 10-meter node range • Assumes no network losses and no queuing delays

46. Topology of Testbed

47. Characteristics of Testbed

48. Unlimited Energy Simulations • First Experiment • All nodes have virtually infinite supply of energy • Ran each data distribution protocol until it converged • Results look at • Data Acquired Over Time • Energy Dissipated Over Time

49. Data Acquired Over Time

50. Data Acquired Over Time