Adaptive Coded Modulation Assisted ISAC Based AFDM Communication In SAGIN Networks

Affine frequency division multiplexing (AFDM) has emerged as a robust multi-carrier modulation candidate for high-mobility communications. This paper investigates an AFDM-based integrated sensing and communications (ISAC) framework for unmanned aerial vehicle (UAV) links within space-air-ground integrated networks (SAGINs). A key contribution of this work is the novel design of the cyclic prefix and postfix (CPP) for AFDM, which is specifically tailored to accommodate wireless power transfer (WPT) requirements, thereby supporting simultaneous information and energy transmission. Specifically, the base station exploits the reflected echoes of AFDM signals to estimate sensing parameters, including the position, velocity, and angle of mobile users. To optimize the communication link, we propose an intelligent adaptive modulation and coding (AMC) decision-making process. A specialized dataset is established, integrating physically interpretable metrics—such as distance, velocity, and angle—with historical AFDM channel state information characterized by its unique chirp-domain representation. Subsequently, a hybrid deep learning architecture, designated as CNN-LSTM, is developed to establish a unified evaluation framework. This framework leverages the feature extraction capabilities of convolutional neural networks (CNNs) to process the spatial-temporal correlations of the AFDM channel, while utilizing Long Short-Term Memory (LSTM) networks to capture the long-term temporal dependencies of UAV trajectories. Simulation results demonstrate that the proposed modeling approach achieves superior separability and robustness, aligning closely with the ideal adaptive envelope while exhibiting enhanced cross-trajectory generalization capabilities compared to conventional methodologies.