Semantic communications prioritize meaning over raw data transmission
Semantic Communications: Principles, Architectures and Applications
Networking and Internet Architecture
Summary
Traditional communication systems focus on sending data accurately as bits, but they don’t check if the information is actually useful for what you want to do with it. This paper reviews a new approach called semantic communication, which cares more about the meaning and relevance of the data for specific tasks instead of just perfect copying. The authors categorize different techniques like deep learning methods and explain how they compare to classic approaches. They also discuss how this approach applies in areas like health devices, self-driving cars, and remote sensing, while pointing out challenges in security and evaluation.
What this means in practice
- •For automotive engineers: Use semantic communication to send only task-relevant data between sensors and control units in autonomous vehicles to improve efficiency and responsiveness.
- •For healthcare iot developers: Design IoT devices that transmit meaningful health data to reduce bandwidth while maintaining quality for monitoring and diagnosis.
A survey. It maps existing work.
Authors
Sasa Maric
Abstract
Conventional communication systems are designed to transmit data reliably, focusing on metrics such as bit error rate, throughput, and channel capacity. However, these traditional measures do not evaluate whether the received information is actually useful for a specific downstream task. To address this limitation, semantic communication has emerged as a promising paradigm that prioritizes the meaning, context, and task relevance of transmitted information rather than raw bit level reconstruction. This paper presents a comprehensive review of the theoretical foundations, system architectures, learning techniques, and evaluation practices in semantic communications. We classify existing state-of-the-art algorithms ranging from deep joint source-channel coding (Deep JSCC) and task-oriented feature transmission to generative diffusion models and contrast their operational trade offs against conventional baselines. Furthermore, we examine key application domains, including remote sensing, autonomous vehicles, and healthcare IoT, while addressing open challenges in security, privacy, robustness, and standardized evaluation.