One
remedy for this is to split the cluster from the most recent merging edge and recluster
each fragment with other clusters in the neighborhood. Re-clustering
can be done recursively until there is no small fragment left. The two control
parameters are N and ?±.
4 Performance Comparison
4.1 Network Model
The overall goal of the simulation is to compare the three clustering schemes in
a sensor network environment.We simulate sensor networks on a 150m?—150m
region. Sensor nodes are randomly deployed with di?®erent levels of density.
The communication range of each sensor is set to 20 meters. Two nodes have
a wireless link between them if they are within the communication range of
each other.
Initially, each node was assigned a unique id, random x, y coordinates
within the region and 0.5J of energy. Then we run the three schemes separately.
For each simulation run, some important statistics were measured.
We assume the same radio model as in [10], where the radio dissipates
Eelec = 50nJ/bit to run the transmitter or receiver circuitry and ?«amp =
100pJ/bit/m2 for the transmit amplifier. We also assume an r2 energy loss
due to channel transmission, where r is the transmission distance. Thus, to
transmit a k-bit message over a distance of d meters using our radio model,
the radio expends
ETx (k, d) = ETx??’elec (k) + ETx??’amp (k, d)
= Eelec ?— k + ?«amp ?— k ?— d2
and to receive this message, the radio expends:
ERx (k) = ERx??’elec (k)
= Eelec ?— k.
Pages:
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568