IPCCC’111 Assessing the Comparative Effectiveness of Map Construction Protocols in Wireless Sensor Networks Abdelmajid Khelil, Hanbin Chang, Neeraj Suri IPCCC 2011
IPCCC’112 Maps Maps are an intuitive data representation technique provide a visual representation of an attribute in a certain area; street map, typographic map, world map, etc. Maps for Wireless Sensor Networks (WSN) applications help users to understand sensed physical phenomena help users to make a decision Sensor locationSensor value (112, 209)145 (218, 163)163 (617, 783)158 (530, 745)163 (477, 625)165 (936, 423)157 (745, 817)155 (653, 237) X Y
IPCCC’113 Sink Map Construction in WSN Naive approach for map construction Energy-efficient approaches for map construction Data collection and processing centrally at sinkin-network Energy efficiency (Comm. complexity on sensor nodes) high comm. overhead Lower comm. overhead Map accuracy node-level accuracy, may decrease because of comm. failures may lose detailed information of each individual node Naive Approach Example of Available Approaches
IPCCC’114 Problem statement and Objectives Several approaches have been proposed. However, Evaluation in carefully selected application scenarios No assessment of the comparative effectiveness of existing approaches: Which is outperforming in Which application/scenario for Which network configuration?
IPCCC’115 Outline Motivation Classification of Existing Map Construction Approaches Performance Comparison in a Wide Range Scenarios Conclusions
IPCCC’116 Data Collection Scheme Classification of Map Construction Approaches Map construction approaches for WSN Region Aggregation Data Suppression Tree-based data collection eScan [9] Isobar [8] Iso-node based data collection Cluster- based data collection Isolines [14] Iso-map [10,11] Contour Map [18] CME [19] Cluster- based data collection CREM [7] Multi-path data collection INLR [16] In-network Processing Technique
IPCCC’117 Region Aggregation Class Basic idea Sensor nodes are ordered hierarchically (clusters, tree..) Every sensor reports to a dedicated node (cluster head, parent..) Dedicated node aggregates adjacent similar data to regions 3 Phases: Region Segmentation At each sensor Non-overlapping polygons Vertex representation Data Collection Aggregator determination Region Aggregation At aggregator Regions formation Aggregation function, e.g. average m m+1 m+2 Tree-based Cluster-basedRing-based
IPCCC’118 Basic idea A subset of sensor nodes (iso-nodes) report their value to the sink suppress similar data to be reported 2 Phases Iso-node Identification what is an iso-node? has a neighbor with different value how to identify? broadcast snoop Isoline Report Generation iso-node based generated at Iso-node routed directly to the sink cluster based generated at cluster-head Iso-node reports to cluster-head a local map Data Suppression Class Isoline Nodes report to the sink Nodes suppress reports to the sink
IPCCC’119 Data Collection Scheme Classification of Map Construction Approaches Map construction approaches for WSN Region Aggregation Data Suppression Tree-based data collection eScan [9] Isobar [8] Iso-node based data collection Cluster- based data collection Isolines [14] Iso-map [10,11] Contour Map [18] CME [19] Cluster- based data collection CREM [7] Multi-path data collection INLR [16] In-network Processing Technique
IPCCC’1110 Selected Map Construction Algorithms The eScan approach [9] Nodes ordered as an aggregation-tree Polygon regions Aggregation function: Average The Isoline approach [14] Local flood to label border nodes Each iso-node reports to the sink Map constructed at the sink [9] Y. Zhao et al. Residual Energy Scan for Monitoring Sensor Networks. In IEEE WCNC, [14] I. Solis and K. Obraczka. Isolines: Energy-efficient Mapping in Sensor Networks. In ISCC, 2005.
IPCCC’1111 Outline Motivation Classification of Existing Map Construction Approaches Performance Comparison in a Wide Range Scenarios Conclusions
IPCCC’1112 Evaluation Framework: Methodology Selected map construction protocols Region aggregation class: eScan Data suppression class: Isoline Simulations using OMNet++ Network Area : 300 x 300 m² Topology: Grid or random Tree-based routing protocol Performance metrics Map accuracy: The ratio of false classified sensors to all sensor nodes. Energy efficiency: Network traffic
IPCCC’1113 Evaluation Framework: Comparative Studies Compare for a wide range of parameters: Impact of physical phenomena properties Hotspot effect range : limited vs. diffusive Hotspot number : 1 vs. n Impact of protocol parameters Sensor value range [0, 60], classes: [0, GV[, [GV, 2GV[... Signal discretization (Granularity value: GV)GV=5…25 Impact of network properties Node densityN=256(16x16) (35x35) Communication failuresBER=0…10 -2 Communication rangeCR=60m
IPCCC’1114 Granularity increases #Isolines and #Iso-nodes decrease -> lower msg overhead Region size increase -> lower msg overhead Accuracy Isoline always outperforms eScan Efficiency Isoline outperforms eScan for lower granularities (b) Step value = 25 unit Comparison: Impact of Granularity (a) Step value = 5 unit
IPCCC’1115 Comparison: Impact of BER BER increases Loss of messages -> lower msg overhead Overhead reduction is higher for eScan Higher BER decreases map accuracy Loss of messages -> gaps in the map Higher accuracy drop for eScan
IPCCC’1116 Comparison: Impact of Node Density Node density increases #Iso-nodes increases -> higher msg overhead #Region and “region border information” increase -> higher msg overhead Node density has low impact on map accuracy Region border precision increases -> provide a more detailed map
IPCCC’1117 Conclusions Region aggregation class Data suppression class + High accuracy with reliable comm. - Less suitable for less reliable comm. + high accuracy for reliable comm. + performs also well for less reliable comm. + accuracy increases with increasing granularity value + Small granularity value + Low density network - Small granularity value + low density network Accuracy Efficiency
IPCCC’1118 Thanks for Your Attention!