Multi-beam scheduling for unmanned aerial vehicle networks

In this paper, we consider the design of an optimum beamforming and scheduling scheme at a multi-antenna UAV central node (UAV-C), which collects information from several other single-antenna distributed UAV nodes (UAV-D). When the channels between UAV-C and UAV-D are both perfectly known, the optimum strategy is to employ minimum mean-square-error (MMSE) beamforming in conjuction with a computationally efficient combinatorial approach. For the case where the channels between UAV-C and UAV-D are subject to channel estimation errors, a stochastic robust design approach is proposed. In this case, it is shown that the optimum strategy is to use robust MMSE beamformers in a combinatorial but efficient manner. Simulation results show that robust beamforming plus scheduling provides significant gains in average throughput over non-robust beamformers with scheduling, and the robust and non-robust beamformers without scheduling. It is observed that, although multi-beam beamforming with scheduling outperforms beamforming without scheduling, the importance of the robust beamformer should not be overlooked as the beamformer and scheduler designed with the assumption of perfect channel knowledge result in significant reduction of throughput in cases where channel estimation is erroneous.