Download presentation
Presentation is loading. Please wait.
Published byMitchell Phillips Modified over 9 years ago
1
University “Ss. Cyril and Methodus” SKOPJE Cluster-based MDS Algorithm for Nodes Localization in Wireless Sensor Networks Ass. Biljana Stojkoska
2
Outline Introduction Localization Techniques –Distributed localization techniques –Centralized localization techniques –Multidimensional scaling Cluster-based localization algorithm Simulation results Conclusion
3
Wireless Sensor Networks (WSN) мно WSN consists of hundreds or thousands of sensor nodes that: sense physical phenomena communicate with each other Why are they so popular? low cost small size easy to install Limitations hardware energy
4
WSN localization Localization is important for: using data gathered from sensor nodes position-aware routing algorithms
5
WSN localization Localization: estimating the location of a node Solution: –installing GPS devices (expensive) –manually (unreliable and inappropriate for many applications) –using algorithmic techniques
6
Distributed localization techniques
7
Trilateration a b c ab c d 2D trilateration 3D trilateration
8
Ad-hoc positioning system 2 1 3 anchordistance Absolute position = 1 2 3 = 4 = 2 (18, 24) 1 1 1 1 2 2 2 2 2 2 3 3 3 3 3 3 3 = 3 * Hop Metric= 6 Niculesu и Nath (2001) Savarese (2002) Savvides (2002)
9
Centralized localization techniques
10
Problem definition Known: a set of N points in a plane coordinates of 0 K < N points (anchors) M N x (N-1) distances between some of the points Should be found: Positions of all N - K points (found unknown coordinates) Abstraction: WSN can be abstracted with graph nodes in WSN ~ vertices in graph distance between nodes ~ edges in weighted graph Analogy: Localization in WSN is analogous with Graph realization (~ find coordinates of the vertices using length of the edges) Semidefinite programming Multidimensional scaling
11
Multidimensional scaling (MDS) is a well known technique used for dimensionality reduction when we have multidimensional data MDS minimize 2 MDS-MAP is an algorithm for nodes localization in WSN based on multidimensional scaling If the distance between nodes i and j can not be measured, it will be approximate with the “shortest path” distance
13
MDS-MAP 1 2 3 4 5 6 7 9 10 11 12 133 8 0 0 0 0 0 0 0 0 0 0 0 0 0 aaa bbb c d e f g h c de f g h c d efg h a+c+d
14
MDS-MAP
15
1 2 3 5 6 7 8 9 10 11 12 133 8 4 1 2 5 6 7 9 10 11 12 13 3 8 4 anchor Linear transform
16
MDS-MAP characteristics Pros –One of the most accurate technique –Relative map creation requires only distances between neighbours –To generate the global map (in 2D) only 3 anchor nodes are needed –The complexity depends on the number of nodes in the network Cons –Centralized processing –Poor accuracy for irregular topologies 1 2 3 4 5 7 9 10 11 12 133 8 a c d 1 2 3 4 5 6 7 9 10 11 12 13 3 8 2
17
грешка =
18
Cluster-based MDS-MAP Aim –To overcome the drawbacks of MDS-MAP –Distributed approach –Improve accuracy for irregular topologies Idea –Divide the network into subsets (clusters) –Apply MDS-MAP on each cluster –Merge local maps into one unique global map Assumptions: –path existence between each pair of nodes in the network –nodes that belong to the same cluster are in close proximity to each other –Each node uses RSSI method for distance estimation –RSSI provide accurate neighboring sensor distance estimation
19
I phase: Initial clustering cluster-head cluster members
20
II phase: Cluster extension gateways
21
III phase: Local map construction
22
MDS-MAP
23
IV phase: Local map merging Referent coordinate system shifting, rotation and reflection of the coordinates Parallel or consecutive merging
24
-Network density (average connectivity of the graph) k=(number_od_edges*2) / number_of_nodes -Number of anchor nodes
25
Simulation results Random and grid based topologies with shape C, L and H Nodes location are obtained using MDS-MAP and cluster- based MDS algorithm (with 5, 7, 10 and 15 clusters) Using different number of anchor nodes (3,4,6 and 10) to generate absolute map Changing radio range, which changes average connectivity of the graph (k, average number of neighbors). 600 different topologies were simulated(6 x 5 x 4 x 5)
26
L topology Random topologyGrid topology MDS-MAP error CB-MDS error random topology grid topology
27
C topology Random topologyGrid topology MDS-MAP error CB-MDS error
28
random topologygrid topology H topology
29
Results discussion Greater connectivity improves the accuracy More anchors improves the accuracy (but not significantly) Number of clusters has a huge impact on the positioning accuracy –In dense graphs (networks), better results can be achieved if the number of clusters is greater –In sparse graphs, the accuracy is greater for small number of clusters
30
Conclusion Which algorithm for nodes localization will be choose depends on: –Desired prediction accuracy –The region where WSN is deployed –The devices’ limitations Cluster-based MDS-MAP is a good solution for: - WSN with irregular topologies - WSN with only a few anchor nodes Cluster-based MDS-MAP as a distributed technique minimize communication cost
31
Any Questions ? THANKS …
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.