Semantic Waveforms for AI-Native 6G Networks

In this paper, we propose a semantic-aware waveform design framework for AI-native 6G networks that jointly optimizes physical layer resource usage and semantic communication efficiency and robustness, while explicitly accounting for the hardware constraints of RF chains. Our approach, called Orthogonal Semantic Sequency Division Multiplexing (OSSDM), introduces a parametrizable, orthogonal-base waveform design that enables controlled degradation of the wireless transmitted signal to preserve semantically significant content while minimizing resource consumption. We demonstrate that OSSDM not only reinforces semantic robustness against channel impairments but also improves semantic spectral efficiency by encoding meaningful information directly at the waveform level. Extensive numerical evaluations show that OSSDM outperforms conventional OFDM waveforms in spectral efficiency and semantic fidelity. The proposed semantic waveform co-design opens new research frontiers for AI-native, intelligent communication systems by enabling meaning-aware physical signal construction through the direct encoding of semantics at the waveform level.