Performance improvements with feedback in cooperative relay networks
Doctor of Philosophy thesis
Recent results on multiple antenna transmission techniques have shown great potential in their ability to improve the overall performance in fading channels. Despite the promise shown by employing multiple antenna's, practical implementations may not be feasible due to size and hardware limitations of mobile nodes. Cooperative Coding is a new transmission paradigm that overcomes these limitations by pooling together the resources of neighboring nodes in a network to create a distributed antenna array. The power of node collaboration can be seen by considering the relay channel, the simplest cooperative network. Recently, protocols have been developed for the wireless relay channel that allow the network to behave as a virtual multiple antenna system. In this thesis we show that in addition to efficient network protocols, exploiting channel state information can yield even more performance in the relay setting by allowing for temporal power and rate control. When power control is used for a given transmission rate, minimizing the outage probability is the appropriate method to maximize performance in the block fading channel. In a relay setting, we derive the optimal power control strategy when the transmitters in the network have perfect knowledge of the network channel state. In practice having perfect channel state knowledge at the transmitters is not possible. In this direction, we derive a power control policy that minimizes the outage probability based on the rate of the feedback link. Interestingly, we observe that only a few bits of feedback are needed to extract much of the gains of the perfect feedback power control policy. For applications that can support a variable rate of transmission, such as data transfers, the feedback can be used to vary both the transmission rate and power. The appropriate performance metric in this case is throughput. We derive throughput maximizing policies for various cooperative transmission protocols. Once again, we show that with a limited rate of feedback, significant throughput gains are possible in relay networks. Interestingly, we show that simultaneous power and rate adaptation is usually not needed. For small average power constraints, power control is imperative, while for large average powers, rate control is sufficient to achieve a large throughput. Our results reveal that power and rate adaptation can lead to significant performance improvements. Even a few bits of feedback can lead to large power savings and throughput gains, and as a result, channel state feedback can be readily implemented with minimal communication overhead in next generation protocols.
Engineering, Electronics and Electrical