Antenna subset selection optimization for large-scale MISO constant envelope precoding

This paper considers robust constant envelope (CE) precoding with antenna-subset selection (AS) in a large-scale MISO downlink scenario where only imperfect channel state information at the transmitter (CSIT) is available. CE precoding is a recently proposed transmission scheme that enables the use of cheap but highly power-efficient power amplifiers, while AS is a well-known approach for reducing the number of power amplifiers. The combination of these two techniques can significantly cut down costs in hardware implementations. We formulate a power minimization problem for AS CE precoding where the worst-case symbol error rate is constrained to be less than a given threshold. The formulation utilizes our recent results on signal characterization of CE precoding. The formulated power minimization optimization problem turns out to be a zero-one linear program. We show that this problem is NP-hard in general. Then, we propose an efficient approximation by Lagrangian dual relaxation and greedy knapsack approximation. Simulation results show that the proposed algorithm can achieve near-optimal performance, and the average number of active antennas accounts for only 19-53% of the total transmit antennas.