Over-the-Air Computation for Partial Aggregation of IoT Data.

The aggregation of a large amount of distributed data is one of the important functions in many IoT applications. Over–the–air computation (AirComp) realizes an efficient aggregation by exploiting the superposition property of wireless medium: all IoT nodes simultaneously transmit their signals, which are received and processed to obtain the aggregated value of their data. The conventional AirComp assumes that all nodes in a given system always join the aggregation process. However, in some practical applications, only a subset of nodes in a given system, which own updated information, are supposed to join the aggregation process. In this case, AirComp should be executed for the partial, unknown subset of nodes, which make it difficult to optimize its transmitting-receiving (Tx–Rx) parameters. To address this issue, in this paper, we propose the over– the–air computation for partial aggregation, called partial AirComp. Our numerical results show that the proposed partial AirComp achieves smaller mean–squared error (MSE) than the conventional AirComp when applied for the partial data aggregation.