Decentralized Learning and Model Averaging Based Automatic Modulation Classification in Drone Communication Systems

Automatic modulation classification (AMC) is a promising technology to identify the mode of received signal in drone communication systems. Recently, benefiting from outstanding performance deep learning (DL), various neural networks (DNNs) have been introduced into AMC methods. Most current methods are based on local framework (LocalAMC) where there only one device, or centralized (CentAMC) multiple devices (LDs) upload their data central server (CS). LocalAMC may not achieve ideal results due insufficient and finite computational power. CentAMC carries significant risk privacy leakage final for training model CS quite massive. In this paper, we propose practical light method decentralized with residual network (ResNet) Simulation show that ResNet-based (DecentAMC) achieves similar while improving efficiency protecting privacy.