Database Reference
In-Depth Information
*01
p
i
*0
p
i
*011
p
i
*00
p
i
*0111
d
nr
*001
p
i
d
r
*0011
*00011
d
l
d
f
p
i
= *011
d(d
r
,p
i
) = 4
d(d
nr
,p
i
) = 1
Learning: (*011%, d
l
)
p
i
*00110
Fig. 3.
Example of Local Learning at Node
n
f
occurred. Simulation results show that the network gains in routing performances
under these conditions.
6 Distances in the W-Grid Logical Network
We also added the Real Distance (RD) feature to W-Grid. Whenever a node
n
j
gets a coordinate from node
n
i
the new coordinate will be one bit longer that
the father one. However
n
i
might have already split and while this information
is known by
n
j
that will know about all the coordinates of it the same is not
for
n
j
's neighbors which are not neighbors of
n
i
. Actually those neighbors could
find useful such kind of information in order to get more precise distance values
during routing. For this reasons routing table entry will also contain this integer
value which represents the real distance among couple of nodes. In section 7 we
evaluated network performance with respect to this feature.
7 Simulation Results
An extensive number of simulations have been conducted in order to evaluate the
network routing performances, in term of average path length, and the robustness
in W-Grid networks.
When comparing W-Grid with GPSR [12], which is the routing method em-
ployed by several existing solutions such as in [3] and in [20], it is appropriate
to remind that in GPSR each sensor needs to be aware both of their physical
location and the network perimeter. These constraints increase the cost of each
sensor and limit GPSR usage possibility, for instance it cannot be used in indoor
environments and in outdoor areas where the density of sensors is beyond the
GPS precision, or when weather conditions are bad.
The simulation model consists of a square area 800
800
m
,inthisarea
205 nodes are randomly spread. Each sensor has its own ID and a radio range
varying from 73
m
to 123
m
(ideal transmission) in order to get different densities,
namely 4
,
8 and 12 neighbors per node respectively. For each scenario we ran 5
×
Search WWH ::
Custom Search