Cooperative Interference Management for Over-the-Air Computation Networks

Recently, over-the-air computation (AirComp) has emerged as an efficient solution for access points (APs) to aggregate distributed data from many edge devices (e.g., sensors) by exploiting the waveform superposition property of multiple (uplink) channels. While prior work focuses on single-cell setting where inter-cell interference is absent, this article considers a multi-cell AirComp network limited such and investigates optimal policies controlling devices' transmit power minimize mean squared errors (MSEs) in aggregated signals received at different APs. First, we consider scenario centralized control. To quantify fundamental performance tradeoff among cells, characterize Pareto boundary MSE region minimizing sum subject set constraints individual MSEs. Though sum-MSE minimization problem non-convex its direct intractable, show that can be optimally solved via equivalently solving sequence convex second-order cone program (SOCP) feasibility problems together with bisection search. This results algorithm computing control, which balances interference-and-noise-induced signal misalignment unique AirComp. Next, other e.g., when there lacks controller. In scenario, introduce temperature (IT) constraints, each constrains maximum total between specific pair cells. Accordingly, AP only needs individually control associated minimization, but IT their neighboring By optimizing levels, shown provide alternative method characterizing same counterpart. Building result, further propose APs cooperate iteratively updating levels achieve Pareto-optimal tuple, pairwise information exchange. Last, simulation demonstrate cooperative using proposed algorithms substantially reduce networks compared conventional approaches.