面向低空经济高动态通信场景的AFDM高效检测器设计OA

In response to the urgent demand for high-dynamic and reliable transmission in airborne communication networks within the context of the low-altitude economy,Affine Frequency Division Multiplexing(AFDM)has attracted significant attention as a waveform technology renowned for its excellent Doppler robustness.However,existing AFDM detectors,whether based on traditional algorithms or neural networks,often face a trade-off between detection performance and computational complexity.To address this challenge,this paper proposes a novel detection algorithm based on an Expectation Propagation(EP)enhanced Graph Neural Network(GNN).The proposed algorithm leverages the inherent sparsity of the AFDM channel matrix to synergistically integrate the EP algorithm with GNN.Specifically,the EP module facilitates an efficient Gaussian approximation of the posterior symbol distribution,while the GNN module employs a sparse pairwise Markov Random Field(MRF)model to capture complex residual dependencies.In addition,a channel approximation strategy is further introduced to significantly simplify the connection structure of the underlying factor graph of the GNN,thereby effectively reducing the overall computational complexity of the EP-GNN detector.Simulation results demonstrate that under doubly-dispersive channel conditions,the proposed EP-GNN detector achieves superior Bit Error Rate(BER)performance compared to conventional detectors and other neural-network-based detectors,while maintaining notably lower computational complexity.These findings underscore the strong potential of the EP-GNN detector for practical applications in highly dynamic low-altitude communication environments within the low-altitude economy.