enac edic 2011 supervisor alcherio...
TRANSCRIPT
MICS
Simulated network topology from a
SensorScope deployment in the Génépi
Mission Increase the accuracy of environmental sensor
networks while decreasing their cost
Accuracy
Can we provide more relevant information by increasing
sampling frequency during interesting events?
Cost
Can we optimize power consumption by suppressing
(spatially or temporally) redundant information?In-network processing allows us to
carefully choose which data we report
Case Study: Constraint Chaining
Overview. Constraint Chaining1 (Conch) is a sensor
network reporting algorithm that uses spatio-temporal
suppression to reduce in-network communication.
Temporal. Stations only transmit sensor values if they
have changed since the last measurement.
Spatial. Stations only transmit a sensor value if its
difference with a particular adjacent station’s value (i.e.,
one that was found to be highly correlated with this
station) has changed.
Adaptation. The network sink performs periodic
optimization to ensure that Conch’s spatial component
takes advantage of patterns in the observed phenomena.
A list of correlated pairs is sent back into the network.
Performance Evaluation
Algorithm Tx reduction Sigma
Uniform sampling 0.0% 0.0%
Temporal suppression 57.3% 6.8%
Constraint Chaining 62.2% 11.5%
Simulated Conch structure from
a SensorScope deployment in
Plaine Morte
Simulation. We have
performed extensive
simulations using real data
from several sensor network
deployments2. Temporal
suppression yields large
energy savings, while
Conch’s spatial component
only manages to bring a
small additional benefit.
Reality. We have delevoped and integrated a power
monitoring board (left) with SensorScope, a commercial
environmental sensing system (right). We are now in the
process of deploying a network of stations on EPFL’s
campus to evaluate Conch under real-world conditions.
References
[1] Silberstein, A., Braynard, R., Yang, J.: Constraint chaining: On energy-efficient
continuous monitoring in sensor networks. ACM SIGMOD 2006, pp. 157–168.
[2] Evans, W. C., Bahr, A., Martinoli, A. Evaluating Efficient Data Collection
Algorithms for Environmental Sensor Networks. DARS 2010.
ENAC /
Performance Evaluation of Data Collection
Algorithms for Environmental Sensor Networks
Author William C. Evans, Alexander Bahr
Supervisor Alcherio Martinoli
All contributors are associated with the Distributed Intelligent Systems and Algorithms Laboratory (DISAL)
EDIC 2011