Distributed MMSE Detection with One-Shot Combining for Cooperative Aerial Networks in Presence of Imperfect CSI

In this paper, we investigate the multiple access technique for the aerial networks, where the multi-antenna base stations are carried by unmanned aerial vehicles (UAVs) to serve ground users. The uplink signals are transmitted by the users simultaneously, then detected and recovered by the aerial networks. On one hand, compared to the case that each UAV individually detects the signals of serving users, we aim to improve the quality of the recovered signals by using cooperative techniques that leverage the signals from multiple UAVs. On the other hand, the fully-centralized cooperation requires signaling exchange between UAVs, which incurs huge signaling overhead and latency, and is infeasible for aerial networks. Therefore, we propose a two-stage distributed minimum mean squared error (MMSE) detection with one-shot signal combining. Specifically, the users' signals are first locally detected by each UAV via the MMSE detector, and then weighted combined at the central UAV in a one-shot manner, where the combining weights are designed to minimize the MSE between the combined signals and the original signals. When only imperfect channel state information is locally available at each UAV, by using the random matrix theory, these combining weights are shown to depend on the long-term channel statistics and thus, greatly reduce the interaction overhead and latency. Numerical results show that the proposed scheme outperforms the non-cooperative detection in terms of the achievable spectral efficiency. Meanwhile, with a much smaller interaction overhead, the proposed scheme achieves comparable spectral efficiency as the fully centralized signal detection.