In our
simulations, the packet losses were mainly due to deadline violation, since each hop drops
the packets which have already expired. The results of Figure 5 indicate, for all the
scenarios under consideration, that scheduling at the application layer by expected
distortion-reduction leads to reduced losses for the most significant classes of packets. This
justifies our use of a scalable video coder that permits such a scheduling. However, each
method achieves different PSNR performance and PLRs depending on its chosen utility
and the presence of network feedback.
As shown in the results of Table 5.2, the ???End-to-end??? case outperforms all other
methods by a significant margin. The ???ETX optimization??? appears to perform relatively
well, even though it is outperformed by approximately 1.5 dB by the ???End-to-end??? case for
Cross-layer Optimized Video Streaming over Wireless Multi-hop Mesh Networks
123
the medium-bandwidth case, and by approximately 1.7 dB for the low-bandwidth case. The
???Localized??? case outperforms the popular ???Highest Bandwidth??? case by approximately 1.1
dB for both medium-bandwidth and low-bandwidth cases, even though the ???Highest
Bandwidth??? case uses full feedback for the status of all the links in each multi-hop
topology.
Pages:
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343