Digital Signal Processing Reference
In-Depth Information
Chapter 7
Distributed Optimization of Local Area
Networks
7.1 Introduction
In Chap. 5, we have demonstrated why the ISM bands can be considered the most
complex wireless scene today. In these bands different heterogeneous networks and
devices co-exist using only LBT etiquettes. The IEEE 802.11 standard is by far the
most popular wireless standard for broadband access operating in the ISM bands. It
is widely recognized that the throughput of IEEE 802.11 networks is dependent on
all of the following parameters:
•
Increasing the
transmission rate
will increase the number of packets that are sent
during a time interval, but more packets can be lost at the receiver due to a lower
interference tolerance.
•
Lowering the
carrier sense threshold
protects a packet against interference, but
decreases transmission opportunities by silencing the transmitter more often.
•
Increasing the
transmission power
will decrease packet losses at the receiver,
but will also reduce spatial reuse in the network.
In this chapter, we introduce a novel control algorithm, Spatial Learning, which
learns the optimal operating point in this 3D design space at RT.
The model of IEEE 802.11 that we have presented in [46], can be used to in-
vestigate the relative importance of the different starvation mechanisms. However,
the assumptions remain restricting for effective RT control. In this contribution, we
have for instance assumed all terminals have the same power, rate and carrier sense
threshold. It is unlikely that in an optimal setting, all terminals will have the same
parameters. Indeed, terminals can be in different scenarios that demand for differ-
ent settings. Hence, if we allow terminals to select their parameters individually the
network throughput can further increase. Also, if we would need to ensure that all
terminals have the same parameters, changes need to be communicated over the
entire network [92]. The large overhead involved reduces throughput and possibly
eliminates the adaptation benefit. Completely relying on DT modeling will hence
not result in an effective RT control. Therefore, we have designed Spatial Learning
relying on the framework presented in Chap. 3. The environment is split up into
