Constraints: Nodes are interdependent. Memory Energy Processing 4/1/2011. Dulanjalie Dhanapala March 30 th, 2011

Transcription

1 PhD. Dissertation Proposal Network Awareness in Wireless Sensor Networks - for Intelligent Organization and Efficient Routing Dulanjalie Dhanapala March 30 th, 2011 ommittee Prof. Anura P. Jayasumana(Advisor) Prof. Indrakshi Ray Prof. Ali Pezeshki Prof. Michael Kirby Introduction -Wireless Sensor Networks (WSNs) Large networks: 1000 to of nodes Sense-pressure, temperature, velocity, humidity, etc. ommunicate wirelessly Nodes are interdependent onstraints: Memory Energy Processing 2 1

2 Background Initially unstructured networks Why structuring/organization required for WSNs? For some applications organization is required Efficient/improved data dissemination Self organization schemes- Localization- GPS/RSSI Requires additional hardware Accuracy issues lustering Virtual oordinate Systems 3 Virtual oordinate Systems - VS Ordinate: Relative position intermsof#hops, w.r.t. an anchor node Ex: A,B,,D anchors A= (0,4,9,6) D= (6,8,5,0) Norm 2 based distance evaluation Geographic information not required Physical voids do not exist in Virtual Domain No additional hardware B= (4,0,5,8) (5,1,4,7) =(9,5,0,5) A= (0,4,9,6) B= (4,0,5,10) D= (6,10,15,0) (5,1,4,11) =(9,5,0,15) 4 2

3 Issues in VS Optimalnumber of anchors and their placement is unknown Affects routing performance Local minima problem Identical coordinates issue Furthest apart anchors No appropriate scheme available to achieves this 5 Problem Statement To develop and characterize techniques and algorithms to achieve reliable and efficient information exchange via self-organization of large WSNs 1. Self organization of wireless sensor networks based on efficient approaches 3. Structuring sensor networks by self learning approaches 2. Methods of extracting topology map, without the need for physical information 4. Making the nodes aware of the network and sensed phenomena 6 3

4 Tasks 1. Self organization of wireless sensor networks based on efficient approaches Regaining lost directionality of VS Directional V based routing and anchor placement Efficient representations of VS 2. Methods of extracting topology map, without the need for physical information Techniques for generating topology maps Topology preserving map based routing Boundary detection without localization 7 Tasks (ontd.) 3. Structuring sensor networks by self learning approaches VS and topology preserving map through learning Reliability, convergence of learning etc. Effect of sleeping cycles, node failures and adding new nodes 4. Making the nodes aware of the network and sensed phenomena Making nodes aware of the network and events/sensed phenomena Accuracy, complexity ost effective implementation 8 4

5 ontributions Regaining lost directionality of VS and anchor placement criterion [1] [5] Topology Preserving Maps (TPM) of 2D and 3D surfaces from VS [1] [7] A= (0,4,9,6) B= (4,0,5,10) D= (6,10,15,0) Hybrid routing Geo-Logical Routing [3] (5,1,4,11) =(9,5,0,15) Novelty of each anchor and compression of Vs [6] Oblivious Sensors to Network- ognizant Smart Sensors [2] Boundary detection using TPM [4] 9 Regaining Lost Directionality of VS Local minima problem Hops to A 1 Norm 2 Distance estimate from each node to destination Hops to A 2 Source (9,3) (8,2) (7,1) A 2 A 1 (6,0) (0,6) (2,8) Destination Transformation to regain the directionality 10 5

6 Directional Transformation in 1D network (9,3) (8,2) (7,1) A 2 A 1 (6,0) (0,6) (2,8) Unit vector in A 1 A 2 direction Vector representation : 11 Directional Transformation in 2D network A 1 A2 2D extension 12 6

7 Angle Between Two Directional Ordinates Two directional ordinates and Angle between i th and j th ordinate directional ordinates Nodes P D where Example: (1) Φ 12 (2) Φ13 Φ 14 (3) (4) 13 Directional Virtual oordinate Routing Step 1: DVS Greedy Forwarding till a local minima Why local minima in DVS? Step 2: How to overcome a local minima Next node is selected, based on a rough estimation of hop distance an design different routing schemes using DVS 14 7

8 R AVG % of SR, LR, GPSR and DVR with 5 randomly placed anchors in spiral, 30 by 30 node grid with 800 nodes, circle with 3 holes, building, and odd networks 5 random anchors, TTL(Time-to-Live) is Anchor Placement Scheme Performance of VS is affected by anchor placement Boundary nodes/furthest apart nodes as anchors ostly localization Nodes at corners are better than edge nodes Alleviate local minima issue Less identical coordinates Distributive identification 16 8

9 Extreme Node Search Extreme nodes are identified as anchors Scatter plot of DVS ordinate under highlighted 17 anchors to identify corner nodes for anchors. Performance Improvement in Routing 18 9

10 Next Regaining lost directionality of VS and anchor placement criterion [1] [5] Topology Preserving Maps (TPM) of 2D and 3D surfaces from VS [1] [7] Hybrid routing Geo-Logical Routing [3] Novelty of each anchor and compression of Vs [6] Oblivious Sensors to Network- ognizant Smart Sensors [2] Boundary detection using TPM [4] 19 Topology Preserving Maps from VS (ntd.) No localization ost effective No additional hardware required Overcome the disadvantages in VS Opens up various opportunities to develop new WSN algorithms 20 10

11 2-D Topology Preserving Maps No location information Nnodes, Manchors P = U S V T VS of Nnodes, Nx M matrix P SVD = US = PV Topological coordinates (Ts), [X T Y T ]= [P SVD (2) P SVD (3) ] 21 ircular network with holes,496 nodes (1) With entire VS (2) Anchors VS High cost Low cost (3) Random nodes VS (4) All the nodes as anchors 22 11

12 343-node Network in a building (1) With entire VS (2) Anchors VS High cost Low cost (3) Random nodes VS (4) All the nodes as anchors 23 Performance Evaluation (ntd.) Network With entire VS 0.90% 1.68% 0.36% 0% 15-Anchors VS 0.83% 1.59% 1.07% 0% 10-Random nodes VS 0.21% 1.50% 0.49% 0% 24 12

13 3D Topology Preserving Maps Two Perpendicular ylinders Each cylinder has a radius 2.54 units, height 16 units, and is covered with a grid of 512 nodes, each with a communication range of random anchors. ylinder with a Hole ylinder of radius 2.54 and height 24 with a hole through it is covered with 490 nodes, each with a communication range of randomly placed anchors 25 Next Regaining lost directionality of VS and anchor placement criterion [1] [5] Topology Preserving Maps (TPM) of 2D and 3D surfaces from VS [1] [7] Hybrid routing Geo-Logical Routing [3] Novelty of each anchor and compression of Vs [6] Oblivious Sensors to Network- ognizant Smart Sensors [2] Boundary detection using TPM [4] 26 13

14 Geo-Logical Routing Algorithm Hybrid routing scheme T -Topology based oordinates V -Virtual oordinates AM - Routes toward closest anchor TTL: time-to-live of a packet 27 Applications: Geo-Logical Routing Avg. Routability GLR-R 94.6% Avg. Routability GPSR 93.8% Avg. Routability GLR-M 98% 28 14

15 Boundary Detection Algorithm If N i has two or less neighbors N i is a boundary node If N i has three or more neighbors: Neighbor triplet N i is a internal node if at least one triplet exists such that 29 2D Network/Event Boundary Detection Dynamic event boundary detection 30 15

16 Boundary Detection in 3D Surfaces 31 Performance Analysis A% (avg) E%

17 Next Regaining lost directionality of VS and anchor placement criterion [1] [5] Topology Preserving Maps (TPM) of 2D and 3D surfaces from VS [1] [7] Hybrid routing Geo-Logical Routing [3] Novelty of each anchor and compression of Vs [6] Oblivious Sensors to Network- ognizant Smart Sensors [2] Boundary detection using TPM [4] 33 Oblivious Sensors to Network- ognizant Smart Sensors Random routing is required in both structured and unstructured network So far, random messages haven t been used for learning Two stages- Virtual coordinates from random routing No initial flooding required Virtual coordinates to Topological coordinates No initial cost 34 17

18 VS Generation From Random Routing Anchors Neighbors [A:2,:3] [A:1] [A:0] B [A:4,B:0] A [A,5,B:1] D [:0] E F 35 Termination riterion of VS Generation Number of packets required for all the nodes to generate error free VS averaged over 10 different configurations as TTL varies. Network has 10,20 and 40 anchors respectively for N=496, 1081 and an use VS based routing Error in VS vs. number of packets forwarded by each node without any V updates, averaged over 10 anchor placements

19 Topology Preserving Maps From Learning Network has VS N13 D N2 B A N27 E N1 ollect source and destination addresses F 37 Topology Preserving Maps From Learning(ntd.) Each node (i) generate basis V, thus can get topological coordinates Q = U S V T P SVD,i = P i V T of node i = V of node i V 2 nd and 3 rd columns of V Topological coordinates based routing Each node develop topology preserving map network awareness 38 19

20 Performance Improvement- Routability 39 Topological Map at a Node-ircular Network with Voids Topology preserving map development at a sample node (identified by color red). Number of anchors is 10 and total number of nodes is 496. TTL of a packet is

21 Topological Map at a Node: T-joint Topology preserving map development at a node of a 3D T-joint assuming it store the destination topology coordinate of every destination node that goes through it. Number 41 of anchors is 10 and total number of nodes is 512.TTL of a packet is 100. Summary Hybrid routing scheme and boundary detection Properties of VS, Novelty analysis and dimensionality reduction Topology map generation from VS for 2D and 3D surfaces Oblivious nodes Network aware nodes Regaining lost directionality and properties of the Directional VS Directional domain routing and effective anchor selection scheme 42 21

22 Future Plans Effect of sleep cycles, node failures and new nodes to VS performance How a node achieves the awareness of sensed phenomena- ompressive sensing approach Oblivious nodes Network aware nodes Beginning of Node/network / phenomena aware systems The effect of error in VS to the performance of network aware system. Accuracy of topology map at a node Topology map improvement by piecewise map generation and stitching together Topology map from Directional Ordinates 43 (a) (b) Topology map from Directional Ordinates (c) (d) (a)network with circular boundary and circular obstacles, with 496 nodes; (b and c)topology map via SVD based method with communication range =1, and 3 (d) Topology maps obtained via Directional Virtual oordinates with only 4 anchors 44 22

23 Suggestions Questions Publications 1. D.. Dhanapala and A.P. Jayasumana, Anchor Placement in WSNs A Directional Ordinate Based Approach, Proc. IEEE global communications conference (GLOBEOM 2011), Dec (to be submitted) 2. D.. Dhanapala and A.P. Jayasumana, lueless Sensors to Network- ognizant Smart Sensors: Topology Awareness Via Self-learning in Wireless Sensor Networks AM SIGOMM 2011, Toronto, Ontario, anada, Aug , 2011(in review) 3. D.. Dhanapala and A.P. Jayasumana, Geo-Logical Routing in Wireless Sensor Networks, Proc. IEEE global communications conference(globeom 2011), Dec (to be submitted) 4. D.. Dhanapala, S. Mehta and A.P. Jayasumana, On Topology Mapping and Boundary Detection of Sensor and Nano-networks Deployed on 2-D and 3-D Surfaces, Proc. 8th Annual IEEE ommunications Society onference on Sensor, Mesh and Ad Hoc ommunications and Networks(SEON), June 2011.(Accepted) 5. D.. Dhanapala and A.P. Jayasumana, Directional Virtual oordinate System for Wireless Sensor Networks, Proc. IEEE International onference on ommunications(i 2011), June 2011.(accepted) 6. D.. Dhanapala and A.P. Jayasumana, Dimension Reduction of Virtual oordinate Systems in Wireless Sensor Networks, Proc. IEEE global communications conference(globeom 2010), Dec D.. Dhanapala and A.P. Jayasumana, Topology Preserving Maps From Virtual oordinates for Wireless Sensor Networks, Proc. 35th Annual IEEE onference on Local omputer Networks (LN 2010), Oct (Best paper award) 45 23