Semantic Communications in Networked Systems: A Data Significance Perspective

Abstract

We present our vision for a departure from the established way of architecting and assessing communication networks, by incorporating the semantics of information, defined, not necessarily as the meaning of the messages, but as their significance, possibly within a real-time constraint, relative to the purpose of the data exchange. We argue that research efforts must focus on laying the theoretical foundations of a redesign of the entire process of information generation, transmission and usage for networked systems in unison by developing (1) advanced semantic metrics for communications and control systems; (2) an optimal sampling theory combining signal sparsity and timeliness, for real-time prediction/reconstruction/control under communication constraints and delays; (3) temporally effective compressed sensing techniques for decision making and inference directly in the compressed domain; (4) semantic-aware data generation, channel coding, packetization, feedback, multiple and random access schemes that reduce the volume of data and the energy consumption, increasing the number of supportable devices. This paradigm shift targets jointly optimal information gathering, information dissemination, and decision-making policies in networked systems.

Understand-Before-Talk (UBT): A Semantic Communication Approach to 6G Networks

“…6, our algorithm outperforms with approximately 40% CPU utilization while learning signatures and the average retrieval time is below 5 ms. Due to aforementioned limitations, and the results in Fig. 6, we find that existing deep hashing are not directly applicable for the SC paradigm (freshness, ageof-information and the extreme low latency requirements of 6G [6]). Using the extended approach developed in Sec.…”

“…Applications that rely on real-time decision-making, such as smart grids and networked control, need a major rethink in their communication processes. There is an increasing interest in the investigation of an efficacious SC system in 6G that can jointly optimize information collection, distribution, and decision-making policies for a large scale autonomous network systems [5], [6].…”

“…While such network systems are not-so-futuristic, the theoretical foundation for SC is poor since "the proper or meaningful piece of information" is neither specified in classical data communication, nor is the process of information generation considered in the communication process. For the last decade, our objective had been to ultra-reliably deliver an entire data stream as quickly as practicable [6]. However, such a classic strategy can indeed be extremely inefficient for 6G systems.…”

“…according to some systems with certain physical or conceptual entities [1]), of communication become significant for 6G [6].…”

“…Fig.6b) shows the comparison of CPU utilization in signature learning and the mean reconstruction time in milliseconds using CIFAR-10 by fixing the signature length to 64 bits. When comparing with other methods[52],[61],[62], in Fig.6, our algorithm outperforms with approximately 40% CPU utilization while learning signatures and the average retrieval time is below 5 ms. Due to aforementioned limitations, and the results in Fig.6, we find that existing deep hashing are not directly applicable for the SC paradigm (freshness, ageof-information and the extreme low latency requirements of 6G[6]). Using the extended approach developed in Sec.…”