18-05-2012, 12:56 PM
Multisensor Data Fusion in Distributed Sensor Networks Using Mobile Agents
10.1.1.10.7251.pdf (Size: 520.96 KB / Downloads: 21)
Introduction
Multisensor data fusion is an evolving technology, concerning
the problem of how to fuse data from multiple
sensors in order to make a more accurate estimation of the
environment [8, 10, 16]. Applications of data fusion cross
a wide spectrum, including environment monitoring, automatic
target detection and tracking, battlefield surveillance,
remote sensing, global awareness, etc. They are
usually time-critical, cover a large geographical area, and
require reliable delivery of accurate information for their
completion.
Mobile agent computing model
Generally speaking, mobile agent is a special kind of
software which can execute autonomously. Once dispatched,
it can migrate from node to node performing
data processing autonomously. Lange listed seven good
reasons to use mobile agents [12], including reducing network
load, overcoming network latency, robust and faulttolerant
performance, etc.
Distributed sensor integration algorithm design
As larger amount of sensors are deployed in harsher
environment, it is important that sensor integration techniques
are robust and fault-tolerant so that they can handle
uncertainty and faulty sensor readouts. Here, the redundancy
in the sensor readouts are used to provide error tolerance.
In this section, we first describe an efficient multiresolution
integration (MRI) algorithm. Then we modify
the algorithm such that the original centralized integration
can be carried out distributively. Readers are referred to
[18] for detailed derivation and case study.
Centralized MRI algorithm
The original MRI algorithm was proposed by Prasad,
Iyengar and Rao in 1994 [17]. The idea essentially consists
of constructing a simple function (the overlap function)
from the outputs of the sensors in a cluster and resolving
this function at various successively finer scales
of resolution to isolate the region over which the correct
sensors lie. Each sensor in a cluster measures the same
parameters. It is possible that some of them are faulty.
Hence it is desirable to make use of this redundancy of
the readings in the cluster to obtain a correct estimate of
the parameters being observed.