4.3 Dynamic Boundary Estimation
The distributed boundary estimation described above is based on an assumption
that the boundary will not change. But in most cases, the target region
may change due to event mobility or faulty sensors. So we need a dynamic
boundary estimation technique to enable working sensors to discover such
changes.
168
Chapter 6 Boundary Detection for Sensor Networks
One such technique is as follows. For each node S, it stores the ID of the
neighboring node that transmitted ID of S. Denote this node as NBD(S);
thus we have NBD(root) = root. At regular intervals, each sensor S compares
its region set to the region set of NBD(S). If the region information does
not match or the node does not get a response from NBD(S), it resets its
region set which only contains its own ID, and passes this information to its
neighbors. So the growing region algorithm would run again, and nodes that
lie in this region would update their boundary set. Of course, the selection of
interval length depends on the balance between accuracy and energy cost.
In [10], the authors deployed six mica2 motes equipped with light sensors
on the roof of a building to detect shadow moving. Their simulation results
are based on scenarios with 100 nodes randomly distributed on a square space
of dimensions 100 ?— 100.
We can get a better performance in terms of accuracy of boundary estimation
in a distributed manner. However, the message passing overhead could be
high, especially in a large area where thousands of sensors are deployed.
Pages:
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288