SENSOR NETWORKS ARCHITECTURE

2. Several thousand nodes Nodes are tens of feet of each other Densities as high as 20 nodes/m3 Sink Internet, Satellite, etc Sink Task Manager SENSOR NETWORKS ARCHITECTURE • I.F.Akyildiz, W.Su, Y. Sankarasubramaniam, E. Cayirci, “Wireless Sensor Networks: A Survey”,Computer Networks (Elsevier) Journal, March 2002.

3. Location Finding System Mobilizer Transceiver Sensor ADC Processor Memory Power Generator Power Unit SENSOR NODE HARDWARE Small Low power Low bit rate High density Low cost (dispensable) Autonomous Adaptive SENSING UNIT PROCESSING UNIT

4. MICA MotesBWN Lab @ GaTech Processor and Radio platform (MPR300CB) is based on Atmel ATmega 128L low power Microcontroller that runs TinyOs operating system from its internal flash memory.

5. UC Berkeley: Smart Dust UCLA: WINS UC Berkeley: COTS Dust JPL: Sensor Webs Rockwell: WINS Examples for Sensor Nodes

6. Rene Mote Dot Mote Mica node weC Mote Examples for Sensor Nodes

7. SENSOR NETWORKS FEATURES • APPLICATIONS: Military, Environmental, Health, Home, Space Exploration, Chemical Processing, Disaster Relief…. • SENSOR TYPES: Seismic, Low Sampling Rate Magnetic, Thermal, Visual, Infrared, Acoustic, Radar… • SENSOR TASKS: Temperature, Humidity, Vehicular, Movement, Lightning Condition, Pressure, Soil Makeup, Noise Levels, Presence or Absence of Certain Types of Objects, Mechanical Stress Levels on Attached Objects, Current Characteristics (Speed, Direction, Size) of an Object ….

8. Factors Influencing Sensor Network Design A. Fault Tolerance (Reliability) B. Scalability C. Production Costs D. Hardware Constraints E. Sensor Network Topology F. Operating Environment G. Transmission Media H. Power Consumption

9. Application Layer Transport Layer Task Management Plane Mobility Management Plane Network Layer Power Management Plane Data Link Layer Physical Layer Sensor Networks Communication Architecture Used by sink and all sensor nodes Combines power and routing awareness Integrates data with networking protocols Communicates power efficiently through wireless medium and Promotes cooperative efforts.

10. WHY CAN’T AD-HOC NETWORK PROTOCOLS BE USED HERE? • Number of sensor nodes can be several orders of magnitude higher • Sensor nodes are densely deployed and are prone to failures • The topology of a sensor network changes very frequently due to node mobility and node failure • Sensor nodes are limited in power, computational capacities, and memory • May not have global ID like IP address. • Need tight integration with sensing tasks.

11. APPLICATON LAYER SMP: Sensor Managament Protocol System Administrators interact with Sensors using SMP. TASKS: • Moving the sensor nodes • Turning sensors on and off • Querying the sensor network configuration and the status of nodes and re-configuring the sensor network • Authentication, key distribution and security in data communication • Time-synchronization of the sensor nodes • Exchanging data related to the location finding algorithms • Introducing the rules related to data aggregation, attribute-based naming and clustering to the sensor nodes

12. APPLICATON LAYER (Query Processing) Users can request data from the network-> Efficient Query Processing User Query Types: 1. HISTORICAL QUERIES: Used for analysis of historical data stored in a storage area (PC), e.g., what was the temperature 2 hours back in the NW quadrant. 2. ONE TIME QUERIES: Gives a snapshot of the network, e.g., what is the current temperature in the NW quadrant. 3. PERSISTANT QUERIES: Used to monitor the network over a time interval with respect to some parameters, e.g., report the temperature for the next 2 hours.

13. APPLICATON LAYER Sensor Query and Tasking Language (SQTL): (C-C Shen, et.al., “Sensor Information Networking Architecture and Applications”, IEEE Personal Communications Magazine, pp. 52-59, August 2001.) • SQTL is a procedural scripting language. • It provides interfaces to access sensor hardware: - getTemperature, turnOn for location awareness: - isNeighbor, getPosition and for communication: - tell, execute.

14. APPLICATON LAYER Sensor Query and Tasking Language (SQTL): • By using the upon command, a programmer can create an event handling block for three types of events: - Events generated when a message is received by a sensor node, - Events triggered periodically, - Events caused by the expiration of a timer. • These types of events are defined by SQTL keywordsreceive, every and expire, respectively.

15. Simple Abtract Querying Example Select [ task, time, location, [distinct | all], amplitude, [[avg | min |max | count | sum ] (amplitude)]] from [any , every , aggregate m] where [ power available [<|>] PA | location [in | not in] RECT | tmin < time < tmax | task = t | amplitude [<|==|>] a ] group by task based on [time limit = lt | packet limit = lp | resolution = r | region = xy]

16. Data Centric Query • Attribute-based naming architecture • Data centric protocol • Observer sends a query and gets the response from valid sensor node • No global ID

17. APPLICATON LAYER Task Assignment and Data Advertisement Protocol INTEREST DISSEMINATION * Users send their interest to a sensor node, a subset of the nodes or the entire network. * This interest may be about a certain attribute of the sensor field or a triggering event. ADVERTISEMENT OF AVAILABLE DATA * Sensor nodes advertise the available data to the users and the users query the data which they are interested in.

18. APPLICATON LAYER Sensor Query and Data Dissemination Protocol Provides user applicatons with interfaces to issue queries, respond to queries and collect incoming replies. These queries are not issued to particular nodes, instead ATTRIBUTE BASED NAMING (QUERY) “The locations of the nodes that sense temperature higher than 70F” LOCATION BASED NAMING (QUERY) “Temperatures read by the nodes in region A”

19. 71 75 68 67 66 71 71 71 68 69 Interest Dissemination Interest dissemination is performed to assign the sensing tasks to the sensor nodes. Either sinks broadcast the interest or sensor nodes broadcast an advertisement for the available data and wait for a request from the sinks. Sink Query: Sensor nodes that read >70oF temperature

20. 68 67 66 Sink 71 71 68 71 69 Data Aggregation (Data Fusion) The sink asks the sensor nodes to report certain conditions. Data coming from multiple sensor nodes are aggregated. 71 75 Query: Sensor nodes that read >70oF temperature

21. Location Awareness (Attribute Based Naming) 71 75 68 67 66 71 71 71 68 69 Query an Attribute of the sensor field Region A Sink Region C Region B Query: Temperatures read by the nodes in Region A Important for broadcasting, multicasting, geocasting and anycasting

22. APPLICATON LAYER RESEARCH NEEDS Sensor Network Management Protocol Task Assignment and Data Advertisement Protocol Sensor Query and Data Dissemination Protocol Sophisticated GUI (Graphical User Interface) Tool

23. Sink TRANSPORT LAYERReliable Multi-Segment Transport (RMST) F. Stann and J. Heidemann, “RMST: Reliable Data Transport in Sensor Networks,”In Proc. IEEE SNPA’03, May 2003, Anchorage, Alaska, USA RMST provides end-to-end data-packet transfer reliability Each RMST node caches the packets When a packet is not received before the so-called WATCHDOG timer expires, a NAK is sent backward The first RMST node that has the required packet along the path retransmits the packet RMST relies on Directed Diffusion scheme RMST Node Source Node

24. Transport Layer PSFQ - Pump Slowly Fetch QuicklyC. Y. Wan, A. T. Campbell and L. Krishnamurthy, “PSFQ: A Reliable Transport Protocol for Wireless Sensor Networks,” In Proc. ACM WSNA’02, September 2002, Atlanta, GA • Packets are injected slowly into the network • Aggressive hop-by-hop recovery in case of packet losses • “PUMP” performs controlled flooding and requires each intermediate node to create and maintain a data cache to be used for local loss recovery and in-sequence data delivery. • Applicable only to strict sensor-sensor guaranteed delivery • And for control and management of the end-to-end reliability for the downlink from sink to sensors • Does not address congestion control

25. Related Work • Wireless TCP variants are NOT suitable for sensor networks • Different notion of end-to-end reliability • Huge buffering requirements • ACKing is energy draining • BOTTOMLINE: Traditional end-to-end guaranteed reliability (TCP solutions) cannot be applied here.  New Reliability Notion is required!!!

26. Reliable EVENT Transport in WSN • NEW NOTION: Reliably Detect/Estimate EVENT features from COLLECTIVE information • Challenges: • Significant energy and processing constraints, multi-hop ad hoc communication • Network congestion Need to address Congestion Control and Reliability in Sensor Networks !

27. Event Radius Sink Sensor nodes Event-to-Sink Reliability O. B. Akan, I. F. Akyildiz, and Y. Sankarasubramaniam, “ESRT:Event-to-Sink Reliable Transport in Wireless Sensor Networks,”in Proceedings of ACM MOBIHOC 2003,pp. 177-188, Annapolis, Maryland, USA, June 2003. Also to appear in IEEE/ACM Transactions on Networking,2004. • Sensor networks are event-driven • Multiple correlated data flows from event to sink • Goal is to reliably detect/estimate event features from collective information • Necessitates event-to-sink collective reliability notion

28. Event Radius Sink Sensor nodes Event-to-Sink Reliability • Sink decides about event features every  time units • Observed event reliability Di , the DISTORTION observed in event estimation in the decision interval i at the sink • Desired event reliabilityD* ,the desired event estimation distortion level for reliable event detection • Application specific, known a priori at the sink • Normalized reliability i =D*/Di • Reporting rate f packet transmissions rate at source nodes

29. Network States

30. ESRTProtocol Overview • Determine reporting frequency f to achieve desired reliability D* with minimum resource utilization • Source (Sensor nodes): • Send data with reporting frequency f • Monitor buffer level and notify congestion to the sink • Sink: • Measures the observed event reliabilityDiat the end of decision interval i • Performs congestion decision based on the feedback from the sources nodes (to determine f >< fmax). • Updates f based on i=D*/Diand f >< fmax(congestion) to achieve desired event reliabilityD*

31. B a f bk bk-1 Db Event ID CN (1 bit) Time Stamp Destination Payload FEC ESRT Congestion Detection Mechanism • ACK/NACK not suitable • We use local buffer level monitoring in sensor nodes bk : Buffer fullness level at the end of reporting interval k Db : Buffer length increment B : Buffer size f : reporting frequency • Mark CN field in packet if congested

32. ESRT OperationFrequency Update

33. ESRT Performance S0 = (NC,LR) S0 = (NC,HR) S0 = (C,LR) S0 = (C,HR)

34. NETWORK LAYER (ROUTING BASIC KNOWLEDGE) The constraints to calculate the routes: 1. Additive Metrics: Delay, hop count, distance, assigned costs (sysadmin preference), average queue length...2. Bottleneck Metrics: Bandwidth, residual capacity and other bandwidth related metrics. REMARK: All routing algorithms are based on the same principle used as in Dijkstra's, which is used to find the minimum cost path from source to destination. Dikstra and Bellman solve the SHORTEST PATH PROBLEM… RIP (Distant Vector Algorithm) -> Bellman/Ford Algorithm OSPF (Open Shortest Path Algorithm)  Dikstra Algorithm

35. Routing Algorithms Constraints Regarding Power Efficiency (Energy Efficient Routing) E (PA=1) F (PA=4) Maximum power available (PA) route Minimum hop route Minimum energy route Maximum minimum PA node route (Route along which the minimum PA is larger than the minimum PAs of the other routes is preferred, e.g., Route 3 is the most efficient; Route 1 is the second). D (PA=3) T Sink A (PA=2) B (PA=2) C (PA=2) Route 1: Sink-A-B-T (PA=4) Route 2: Sink-A-B-C-T (PA=6) Route 3: Sink-D-T (PA=3) Route 4: Sink-E-F-T (PA=5)

36. Why can’t we use conventional routing algorithms here? Global (Unique) addresses, local addresses. Unique node addresses cannot be used in many sensor networks • sheer number of nodes • energy constraints • data centric approach Node addressing is needed for • node management • sensor management • querying • data aggregation and fusion • service discovery • routing

37. NETWORK LAYER (ROUTING for SENSOR NETWORKS) Important considerations: • Sensor networks are mostly data centric • An ideal sensor network has attribute based addressing and location awareness • Data aggregation is useful unless it does not hinder collaborative effort • Power efficiency is always a key factor

38. Some Concepts • Data-Centric • Node doesn't need an identity • What is the temp at node #27 ? • Data is named by attributes • Where are the nodes whose temp recently exceeded 30 degrees ? • How many pedestrians do you observe in region X? • Tell me in what direction that vehicle in region Y is moving? • Application-Specific • Nodes can perform application specific data aggregation, caching and forwarding

39. Taxonomy of Routing Protocols for Sensor Networks Categorization of Routing Protocols for Wireless Sensor Networks: (K. Akkaya, M. Younis, “A Survey on Routing Protocols for Wireless Sensor Networks,” Elsevier AdHoc Networks, 2004) 1. Data Centric Protocols Flooding, Gossiping, SPIN,SAR(Sequential Assignment Routing), Directed Diffusion, Rumor Routing, Gradient Based Routing, Constrained Anisotropic Diffused Routing, COUGAR, ACQUIRE 2. Hierarchical LEACH, TEEN (Threshold Sensitive Energy Efficient Sensor Network Protocol), APTEEN, PEGASIS, Energy Aware Scheme 3. Location Based MECN, SMECN (Small Minimum Energy Com Netw), GAF (Geographic Adaptive Fidelity), GEAR

40. B D G C A E F Conventional ApproachFLOODING Broadcast data to all neighbor nodes

41. ROUTING ALGORITHMS Gossiping GOSSIPING: Sends data to one randomly selected neighbor. Example:

42. Problems of Flooding and Gossiping PROBLEMS: Although these techniques are simple and reactive, they have some disadvantages including: * Implosion (NOTE: Gossiping avoids this by selecting only one node; but this causes delays to propagate the data through the network) * Overlap * Resource Blindness * Power (Energy) Inefficient

43. (a) (a) A Implosion B A C B (a) (a) D C q s (r,s) (q,r) Problems Data Overlap r • Resource Blindness No knowledge about the available power of resources

44. Gossiping • Uses randomization to save energy Selects a single node at random and sends the data to it • Avoids implosions • Distributes information slowly • Energy dissipates slowly

45. B D G C A E F The Optimum Protocol • “Ideal” • Shortest-path routes • Avoids overlap • Minimum energy • Need global topology information

46. SPIN: Sensor Protocol for Information via Negotiation(W.R. Heinzelman, J. Kulik, and H. Balakrishan, “Adaptive Protocols for Information Dissemination in Wireless Sensor Networks”,Proc. ACM MobiCom’99, pp. 174-185, 1999 ) • Two basic ideas: • Sensors communicate with each other about the data that they already have and the data they still need to obtain • to conserve energy and operate efficiently • exchanging data about sensor data may be cheap • Sensors must monitor and adapt to changes in their own energy resources

47. SPIN Good for disseminating information to all sensor nodes. SPIN is based on data-centric routing where the sensors broadcast an advertisement for the available data and wait for a request from interested sinks 1. 1. ADV 2. REQ 3. DATA 2. 3.

48. ADV DATA REQ ADV REQ DATA SPIN

49. ROUTING ALGORITHM (DIRECTED DIFFUSION) (C. Intanagonwiwat, R. Gowindan and D. Estrin, “Directed Diffusion: A Scalable and Robust Communication Paradigm for Sensor Networks”, Proc. ACM MobiCom’00, pp. 56-67, 2000.) • This is a DATA CENTRIC ROUTING scheme!!!! • The idea aims at diffusing data through sensor nodes by using a naming scheme for the data. • The main reason behind this is to get rid off unnecessary operation of routing schemes to saveEnergy. Also Robustness and Scaling requirements need to be considered.

50. Gradient Setup Data Delivery Interest Propagation Directed Diffusion Source Sink