Multiple-access block-fading channels

n this work we analyze the effect of delay constraints on the performance of multiple-access fading channels. In this contest, due to the slow fading dynamics with respect to the tolerable decoding delay, only a finite number of independent fading states affect each codeword. We study the effect of delay constraints from two different
perspectives.
In the first part of the work, we consider a simple, decentralized,uncoordinated system where users access at random the channel whenever they have data to transmit. To cope with background noise, fading andinterference from other users, packets that are negatively acknowledged by the receiver are retransmitted. At the receiver, packets related to the same information bits are combined together to increase decoding reliability. We study three different protocols (a generalized version of Aloha, a repetition protocol with maximal ratio combining at the receiver and an incremental redundancy protocol) in terms of total throughput (bit/s/Hz) as function of several different parameters and then we study its limiting behavior with respect to those parameters.
Then, we perform a comparison between systems that implement at MAC layer one the above protocols and at physical layer different decoding strategies. We compare the optimized throughput as a function of the average transmit energy per bit. In doing such a comparison we get insight into the optimal choice of the the system parameters, astransmission rate and average channel load.
In the second part, we consider a completely centralized system, where users know the channel state and can vary rate and power according to the channel conditions so that their rate is inside the fading dependent achievable rate region. We define a variable coding scheme and study the corresponding long-term average capacity region. Since we assume that codewords can span a maximum number of consecutive time slots and that the channel is know only up to the current slot, the optimal solution is given in terms of Dynamic Programming algorithm.
Then, we consider the wideband performance of the variable rate coding scheme described above. We characterize both the performance in the infinite bandwidth regime, i.e., the minimum transmit energy per bit, and in the wideband regime, i.e., wideband slope region. We show that the simple ``one-shot'' policy that concentrates all the energy in only one of the fading states, is wideband optimal.