On the Energy Efficiency of Cooperative Wireless Networks

The aim of this dissertation is the study of cooperative communications in wireless networks. In cooperative networks, each user transmits its own data and also aids the communication of other users. User cooperation is particularly attractive for the wireless medium, where every user listens to the transmission of other users. The main benefit of user cooperation in wireless networks is, probably, its efficacy to combat the wireless channel impairments. Path loss and shadowing effects are overcome using intermediate nodes, with better channel conditions, to retransmit the received signal to the estination. Further, the channel fading effect can be also mitigated by means of cooperative spatial diversity (the information arrives at the destination through multiple independent paths). These benefits result in an increase of the users spectral efficiency and/or savings on the overall network power resource. Besides these gains, the simple idea of cooperation expands enormously the communication possibilities, compared to classical communications. For instance, in cooperative networks the interference caused by a source terminal on its neighbor nodes can be seen as a useful signal and used to aid the communication of other nodes. Cooperation also changes the classical idea of the channel as a simple link between a source and a destination. If users cooperate, then any node that serves as a relay becomes an element of the channel, just as reflecting obstacles that cause signal fading. Thus, the channel is no more one link but the network itself. Due to these new possibilities, the design of cooperative networks has motivated new problems at all the levels of the communication protocol stack. In this dissertation, we address these problems by analyzing the energy efficient regime of different cooperative communication systems. In particular, we are interested in studying, the spectral efficiency as a function of the transmitted power per information bit relative to the noise spectral level. Obtaining the spectral efficiency for all values of energy per bit is usually unfeasible. Instead, if the energy efficient regime coincides with the low power regime, the communication strategy can be well characterized by computing two fundamental metrics: the minimum energy that we need to dedicate to each transmitted bit to have a reliable communication or, equivalently, the maximum rate that can be achieved per unit energy (RPE) and also the slope of the spectral efficiency at the point of minimum energy per bit. This slope indicates the bandwidth efficiency. Throughout the dissertation it is assumed an average total network power to be shared with all users. According to the channel magnitudes and phases, which we assume constant and available at all network nodes, the total power is optimally allocated among users and transmission intervals to maximize the RPE and the slope of the spectral efficiency. We consider three basic cooperative channels: the single relay channel, the two-user cooperative multiple access channel, and the multi-hop multiple relay channel. This channels capture the essence of user cooperation and serve as primary building blocks for cooperation on a larger scale. Even for the simplest one, the relay channel, the capacity is not known today. Therefore, we mainly focus on studying spectral efficiencies achievable with decode and forward (DF) protocols and capacity upper bounds derived with the cut-set bound. In these cases, the low power analysis is a useful tool to study the energy efficiency of the communication system. Firstly, we study the single relay channel, where a source communicates to a destination aided by a dedicated relay. For this channel, we analyze the benefits associated with different capabilities at terminals, such as: i) the phase synchronization between the source and the relay, that allows terminals to transmit signals that add coherently at the destination; ii) the full duplex (FD) capability at the relay, that allows the relay to receive and transmit simultaneously in the same band; and iii) the channel access via superposition, that jointly with a receiver able to cope with inter-user interference, allows the source and the relay to transmit simultaneously. The synchronism benefit can be observed by computing the maximum RPE. If the relay can not work in FD mode, the relay works in half duplex (HD) mode and the transmission and reception channels are orthogonal, e.g. time-division; Likewise, if superposition channel access is not possible, then the transmission must be orthogonal. In both situations, the resultant bandwidth inefficiencies can be observed by computing the slope of the spectral efficiency. For the relay channel, we also study the energy efficient regime using other forwarding protocols, such as amplify and forward or compress and forward, and extend the energy efficiency analysis to ergodic fading channels, in order to assess the impact of the channel fading statistics. Secondly, we investigate the two-user cooperative multiple access channel, where two nodes cooperate with each other in transmitting information to a common destination. The gains provided by the same terminal capabilities considered for the relay channel are revisited here. In addition, we study the gains provided by jointly coding, via superposition, the own generated data and the cooperative data, instead of transmitting them as separated data flows. We design new coding schemes to accommodate the HD/FD modes, the superposition/orthogonal channel access and the joint coding and data flow separation. Finally, we consider the extension of the relay channel to the multi-hop multiple-relay channel, where a source communicates with a destination aided by several relay nodes that listen to all the transmissions. In this case, we restrict the analysis to asynchronous and orthogonal transmissions. Although we only consider DF-like protocols, several coding schemes are possible depending on: i) whether all the relays decode the source message (allcast) or only the destination (unicast and depending on ii) whether nodes use multiple received signals to decoded the message (accumulative) or only one signal (non-accumulative). For each of these possibilities, we design a different multi-terminal coding strategy and compute the maximum RPE. By maximizing the RPE, we provide a joint solution to a set of problems traditionally belonging to different layers: power allocation (physical), relay selection, and routing (network). Among all the possibilities, we find solutions for distributed scenarios.