Semantic Non-Terrestrial Communications With Open RAN-Enabled 6G

<p>The integration of semantic communication with<br>Open Radio Access Network (O-RAN)-enabled Non-Terrestrial<br>Networks (NTN) is a key enabler for 6G, optimizing band-<br>width efficiency, Artificial Intelligence (AI)-native inference and<br>intelligent satellite-terrestrial integration. Unlike traditional bit-<br>level transmission, semantic-aware networks extract and transmit<br>only meaningful information, reducing overhead and improving<br>the adaptability. This paper presents SEM-NTN, a semantic-<br>aware O-RAN-enabled NTN framework, that leverages AI-<br>driven feature extraction, adaptive compression, and dynamic<br>resource allocation. Simulation results show that SEM-NTN<br>reduces transmission delay by up to 87.5% while maintaining<br>AI inference accuracy close to that of full-quality compression.<br>Notably, semantic-aware compression achieves up to 84.6% mean<br>Average Precision (mAP) and 77.6% mean Intersection over<br>Union (mIoU) under a 5ms delay constraint–closely matching the<br>uniform baseline (85% mAP, 78% mIoU) but with significantly<br>improved latency performance. These findings highlight SEM-<br>NTN’s potential for scalable, low-latency, and resource-efficient<br>6G communications. The paper concludes by outlining key<br>challenges in semantic protocol design, real-time adaptation, and<br>standardization for AI-native network integration.</p>