Multi-terminal modulation classification network with rain attenuation interference for UAV MIMO-OFDM communications using blind signal reconstruction and gradient integration optimization

Abstract

The field of Automatic Modulation Classification (AMC) has emerged as a critical component in the advancement of next-generation intelligent Unmanned Aerial Vehicles (UAVs), 6G cognitive space communications, and spectrum regulation initiatives. Our research introduces an innovative AMC algorithm tailored for UAV MIMO-OFDM communication systems. This algorithm leverages blind signal reconstruction, constellation density matrix analysis, multi-terminal decision fusion, and model optimization training to enhance performance. The algorithm begins with the application of blind source separation to reconstruct signals and bolster their representation capabilities. Subsequently, we introduce a novel feature, the Enhanced Constellation Density Matrix (CDM), crafted to withstand the challenges posed by UAV channel interferences while providing a robust representation of the constellation diagram. Building upon this foundation, we propose the UAV-Decision Fusion Network (UAV-DFNet), an advanced network that utilizes CDM features as inputs to deeply mine signal characteristics and achieve superior signal recognition accuracy. To further refine the classification precision, we implement dual strategies: multi-terminal decision fusion and gradient integration, into the UAV-DFNet. Comprehensive experimental results substantiate the effectiveness and superiority of our UAV-DFNet classifier over existing deep learning (DL)-based classifiers, demonstrating its potential to significantly advance the state of the art in UAV cognitive communications and beyond.