Semantic Communication for Multimodal Machine Services under Imperfect Channel State Information, Heterogeneous Quality-of-Service Demands, and Resource Constraints in Integrated Sensing and Communication Networks

Abstract

The evolution of next-generation wireless networks has catalyzed a paradigm shift from traditional Shannon-based data transmission toward semantic-aware communication architectures. This transition is particularly critical for multimodal machine services operating within Integrated Sensing and Communication networks, where massive volumes of heterogeneous data must be transmitted under strict latency and reliability constraints. However, realizing the full potential of semantic communication in these environments is severely hindered by imperfect channel state information, diverse quality-of-service demands, and stringent computational and spectral resource limitations. This paper provides a comprehensive investigation into the design, optimization, and evaluation of a robust semantic communication framework tailored for multimodal machine-type communications in Integrated Sensing and Communication systems. By treating the semantic extraction and resource allocation processes as a joint optimization problem, this study analyzes how varying degrees of channel state information uncertainty impact semantic fidelity and overall system utility. Furthermore, the research explores the dynamic management of heterogeneous quality-of-service requirements across different machine tasks, ranging from real-time autonomous control to high-resolution environmental monitoring. Through extensive theoretical analysis and systematic evaluations, the study elucidates the complex trade-offs between sensing resolution, semantic compression efficiency, and transmission robustness. The findings offer critical insights into developing adaptive, resource-efficient, and resilient communication protocols that can sustain advanced machine intelligence in highly dynamic and resource-constrained wireless environments.