Channel reconfiguration distributed interference alignment for multi-unit over-the-air computation

Abstract

Over-the-air computation (AirComp) has been widely studied as an emerging technique for efficient data aggregation. In multi-unit AirComp networks, inter-unit interference seriously affects the data aggregation performance. To solve the above problem, in this paper, a novel scheme of channel reconfiguration distributed interference alignment (CRDIA) is proposed. We virtualize several devices within a unit as one transmitting device, reconstruct the channel matrix of the system, and iterate iteratively design the pre-processing and post-processing components. Based on the above, the pre-processing matrices of AirComp devices and the post-processing matrices of fusion centers are designed. The sum rate of the system is used as a metric for interference cancellation and mean-squared error (MSE) is used as a metric for data aggregation accuracy. We evaluate the proposed scheme by simulating a network with various numbers of AirComp devices and units. The simulation results show that the CRDIA scheme can achieve accurate computation of the AirComp objective function without inter-unit interference.