Energy efficiency optimization in the multiantenna downlink with linear transceivers

Optimization of transmit strategies with linear transceivers in multiple-input multiple-output (MIMO) broadcast channels generally leads to nonconvex problems, which cannot be solved efficiently in a globally optimal manner. Instead, it is necessary to resort to suboptimal algorithms. In this paper, we evaluate the application of a gradient descent algorithm for the optimization of the energy efficiency in such a system. Since the quality of the obtained locally optimal solutions depends on the initialization, a successive stream allocation is introduced and combined with the gradient algorithm. Comparison with a globally optimal reference algorithm for the special case of single receive antennas shows that the obtained solutions are close to the global optimum. For the MIMO case, the energy per bit achievable with dirty paper coding, which is a lower bound for the case of linear transceivers, is used as benchmark, and good performance of the gradient-based methods is shown for MIMO systems as well.