Non-Cooperative Compute-and-Forward Strategies in Gaussian Multi-Source Multi-Relay Networks

Compute-and-forward is an emerging technique to increase throughput in interference networks. It allows the relays to decode an equation of codewords instead of each single codeword. Choosing the appropriate equation plays a crucial role on the performance. While there are several algorithms that solve this optimization problem, they all perform a local optimization. In large relay networks this can result in linear dependent equations at the destination and a large outage probability. In this paper we propose new non-cooperative strategies on how to choose the desired equations and enforce linear independent equations. This allows to solve the optimization problem in a localized and distributed manner. Further we compare the achievable sum-rate of those strategies and show that correlated signals have a huge impact on the performance.