Semantic-Based Channel State Information Feedback for AAV-Assisted ISAC Systems

Abstract

For autonomous aerial vehicles (AAV)-assisted integrated sensing and communication (ISAC) systems, a semantic-based channel state information (CSI) feedback scheme is proposed in this article. Unlike traditional full CSI feedback, the proposed scheme minimizes feedback burden by utilizing predefined semantic databases at both the transmitter and receiver. First, a deep-learning-based clustering method is developed to construct the semantic database from measured CSI samples. Then, an incremental clustering-based identification method is proposed, enabling dynamic updates and adjustments to semantic databases as new CSI is continuously acquired. Finally, the proposed CSI feedback scheme is validated through scenario identification, and extensive channel measurements are conducted in three typical campus scenarios: 1) playground; 2) lake; and 3) buildings. The results show that the accuracy of the semantic feedback-based scenario identification reaches 97.5%, which is 0.6% higher than the accuracy of the full-CSI feedback-based scenario identification. Specifically, the CSI is fed back through semantic database labels, requiring only a few bytes. This significantly reduces feedback burden while maintaining high accuracy of ISAC tasks. Furthermore, the proposed feedback scheme can also be extended to other AAV-assisted applications, such as the Internet of Things and emergency response.