Download presentation
Presentation is loading. Please wait.
Published byBarnard Fleming Modified over 8 years ago
1
ASAP: An Adaptive Sampling Approach to Data Collection in Sensor Networks Bugra Gedik Ling Liu Philip S. Yu IEEE Transactions on Parallel and Distributed Systems 2007 1
2
Introduction Sensors are power constrained Event-based data collection Periodic data collection Model-based data collection Intra-node modeling Inter-node modeling 2
3
System Model (1/2) 3
4
System Model (2/2) Adaptive data collection and model prediction 1. Collect sensing data for updating 2. Perform prediction Correlation-based sampler selection 1. Divide nodes into sub clusters 2. Select sampler nodes Sensing-driven cluster construction 1. Cluster formation 2. Cluster head selection 4
5
Sensing-driven Cluster Construction (1/3) 5
6
Sensing-driven Cluster Construction (2/3) m.org m.ttl m.rnd m.src m.dmu Hop distanceData distance 6
7
Sensing-driven Cluster Construction (3/3) 1.A limited TTL 2.Cluster-connection tree 7
8
Correlation-based Sampler Selection (1/3) 8
9
Correlation-based Sampler Selection (2/3) 9
10
Correlation-based Sampler Selection (3/3) 10
11
Model-based Prediction A. Prediction Model The values of non-sampler nodes are predicted by probabilistic model Forced samples do not propagate up to the based node Non-sampler Sampler 11
12
Model-based Prediction 12
13
Evaluations 13
14
A. Messaging Cost 14
15
B. Energy Consumption 15
16
C. Data Collection Quality(1/3) 16
17
C. Data Collection Quality(2/3) 17
18
C. Data Collection Quality(3/3) Only messaging and sensing consume energy Negative MAD implies the network exceeds its lifetime 18
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.