Even though antennas may be placed with different
orientations, in the optimal solution they are mounted perpendicular to the walls.
Figure 10.9 displays results from different search resolutions ranging from 0.1 to 0.4 m
that were used to ?¬?nd the best antenna placements in the example, but with the coverage
reevaluated with the ?¬?nest discretization resolution used (0.1 m). In all cases, M??450
discretized orientations were used for read accuracy calculations. The coverage percentage
is represented by the proportion of discretized points in the 23232 m portal space that
can be read with 90% read accuracy. It can be seen that coarser search resolutions may or
may not ?¬?nd the same solution as a ?¬?ner search resolution. When 0.3 or 0.4 m resolution is
used, the enumeration method results in suboptimal solutions.
To evaluate the effect of M, a value of M??1916 is ?¬?rst chosen for calculating the read
accuracy of each point, and then the calculation is repeated with smallerMvalues. Table 10.1
shows the extent of the two types of errors that are caused by smaller M values.
The two types of errors stabilize and quickly converge to a very small value as M
increases. In particular, the total classi?¬?cation error is well below 1% once M reaches a
value of 450. This point is further illustrated by Figure 10.10, which shows the percentage
of tag locations that achieve 90% read accuracy coverage (at the ?¬?nest tag space discretization)
using the optimal solution that is determined by each value of M.
Pages:
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385