Digital Signal Processing Reference
In-Depth Information
at physical layer to the scheduling period at medium access control layer [38]. De-
pending on the exact degree of the dynamics (i.e., coherence time), different adapta-
tion strategies are optimal. For instance, when time variations can be tracked timely,
it is optimal to adapt accordingly. However, when variations are too fast, diversity
techniques are very powerful. It is however widely accepted that the coherence time
for indoor and stationary users can easily be larger than 100 ms (see Chap. 6). This
implies that channel variations are fast, but it is still possible to track these variations
in the lower layers.
Besides variations in time expressed by the coherence time, channels are typi-
cally also varying over frequency. This variation is captured in the coherence band-
width, and frequency-adaptive techniques exist [37] in case the communication
channel bandwidth is larger than the coherence bandwidth.
Finally, the channel is varying in space. This last variation means that different
users in the same network often see a very different channel since they are located
in a slightly different spot. Furthermore, the use of multiple antennas on a single de-
vice exactly exploits the fact that both antennas see a different channel since they are
separated in space [39]. Variations over space will be exploited to improve the mon-
itoring in Chap. 4 and to make the RT procedure more energy efficient in Chap. 6.
Solutions should hence be robust to the channel variations in time, frequency and
space.
3.3.2.2 Application Dynamics
Networked applications communicate in packet bursts that are transmitted over the
communication network. Such bursty behavior is typically more difficult to track
and to predict compared to more continuous variations. Next to packet burstiness
within application streams, the average bitrate requirements can differ a lot from
application to application.
Different applications are also sensitive to different performance aspects. Real-
time streaming applications prefer strict delay guarantees for a frame, while file
transfer is best served by maximizing the average throughput, hence total file down-
load delay. Sensor networking applications are typically not throughput intensive.
For those applications, delay and robustness against failure are the most important
performance aspects.
3.3.2.3 Network Dynamics
Mobile networks do not necessarily rely on an extended infrastructure to operate and
can be established on the fly, i.e., in an ad-hoc way. Alternatively, in infrastructure
mode, the network changes as terminals join and leave the network. As a result,
devices can take part in networks of virtually any size or topology, and the number
of users in the network will vary over time. Solutions should therefore be able to
optimize networks of different size or topology. Network dynamism is limited by
