Optimizing soft information in relay networks

We classify relay forwarding schemes on the basis of soft and hard information. While soft information is highly beneficial, it may also incur additional expenditure of network resources such as power and bandwidth. The relay functionality optimization problem is therefore related to optimizing soft information. With SNR as a metric, we show that the optimal soft information relay mechanism for uncoded communication is the estimate and forward (EF) policy. We demonstrate that EF is well suited to both parallel and serial relay networks due to its optimized soft information content. We also provide an example of a hybrid network where the performance gain from EF is significant.