Digital Signal Processing Reference
In-Depth Information
different scenarios and for each scenario a DT policy is built. However, to be de-
tectable the scenarios used in this chapter cover a large amount of run-time situation
and the optimal strategy for each run-time situation is not the same, as it's highly
dependent on the situation of its neighbors. Therefore, learning becomes necessary
to calibrate the DT policy for the observed scenario to optimally fit the experienced
run-time situation.
Our novel algorithm exhibits the following desired properties for an IEEE 802.11
control algorithm [93]. It's a single-channel solution and maintains interoperability
with other standards. Furthermore it is backward compatible with all IEEE 802.11
protocols as long as a terminal has a tunable power, rate and carrier sense threshold
selection.
A lot of work has been done on optimizing these control parameters individu-
ally [94-96]. More recently, the optimization of two parameters at once has been
investigated [97, 98]. To the best of our best knowledge, this chapter presents one
of the first algorithms that tries to optimize all three parameters simultaneously.
This chapter is structured as follows. First, we present related work in Sect.
7.2
.
Our benchmark solution, Spatial Backoff, is explained in more detail in Sect.
7.2.2
.
We also explain the possibilities, remaining for future control algorithms. Our con-
trol algorithm, Spatial Learning, is discussed in Sect.
7.3
. Simulation results are
presented in Sect.
7.4
. Finally, we present our concluding remarks in Sect.
7.5
.
7.2 Existing Flexibility and Control Mechanisms
In this section we present some of the recent work on IEEE 802.11 optimization and
present a general overview of multi-agent learning, i.e., when multiple terminals
learn simultaneously.
7.2.1 Optimization of IEEE 802.11 Networks
Here, we present an overview of different algorithms that exploit the flexibility of
transmission rate, carrier sense threshold or transmission power.
7.2.1.1 Transmission Rate
The main goal of rate adaptation algorithms is to determine the current channel state
and select the optimal data rate accordingly.
One of the earliest rate adaptation algorithms is called Automatic Rate Fallback
(ARF) [99]. Here, the rate is increased after
S
consecutive successful transmissions
and dropped after
F
consecutive failures. ARF was improved by dynamically tun-
ing
S
and
F
with Binary Exponential Backoff [100]. This avoids overprobing high
rates and overutilization of low rates. Receiver-Based AutoRate (RBAR) monitors
