• link adaptation
• channel prediction
• HARQ

A promising way of improving the spectral efficiency of radio interfaces beyond
3G is by performing fast link adaptation. Since the channel is fluctuating due to
fading and shadowing effects, the transmission parameters of a wireless communication
system have to be adjusted according to the current channel state in order
to make best use of the available resources. In real-world systems, anyway, only
partial channel information is available at the transmitter. This has a strong impact
on achievable performances since, without exact knowledge of the fading state, it
is no longer possible to accurately adapt coding and modulation. The main objective
of this thesis is to design a link adaptation scheme which is robust despite
of this uncertainty about the channel. To this end, in a preliminary phase it is at
first quantified the performance loss with respect to the ideal case of complete information
and consequently it is studied how it can be minimized with the use of
channel prediction. Since a severe degradation can still be observed in some fading
conditions, an additional mechanism for fast link-level retransmissions based
on soft-combining of information is suggested. The retransmission protocol is used
not just to recover from incorrectly received codewords, but the link adaptation
strategy is defined anew so as to enhance throughput with the new degree of freedom
represented by delay. With this perspective, we get closed-form expressions
of the interesting parameters for the case of block fading and Chase combining, whilst a remarkable improvement of the outage probability approximation of the
word error rate for the case of incremental redundancy and/or selective channels is
also presented. With this analytical framework an optimization of the adaptive system
for any criterion and constraints is thus feasible at low computational cost without
necessarily resorting to time-consuming link-level simulations. In the examples
provided, encouraging results are obtained since it is observed that the achieved
throughput comes relatively close to the throughput of an ergodic fading channel at
the expense of few retransmissions, even when no channel state information apart
from the long-term average SNR is available at the transmitter.