Precoder selection scheme based on message- passing approach: a practical perspective

This work addresses a distributed technique to precoding matrices coordination in a multi-cell network. We propose an iterative method based on a message-passing algorithm in factor graphs. Finite precoder codebooks for different MIMO setups are adopted for beam forming purposes. Conversely, practical cellular networks usually solve the precoder selection problem in a non-iterative manner based on the best response method. Nevertheless, this method may not provide a near-optimal solution. For more realistic results, the wireless channel is modeled based on measured data that comprises spatial correlation effects. Evaluations on the potential of such an approach are provided and its performance is compared with the selfish/greedy approach. The capability of reaching the globally optimal solution is evaluated, as well as its performance per iteration. Simulation results for the precoder selection example are presented and discussed, which show that the graph-based technique generally obtains gain in system capacity over both the non-iterative and selfish approaches. Besides, the proposed method usually reaches the global optima in an efficient manner in terms of signaling load.