Semantic Communications for Simultaneous Wireless Information and Power Transfer

Abstract

In this paper, we study the fundamental limits of simultaneous semantic information and power transfer in wireless networks, where we consider both the point-to-point case as well as the Gaussian multiple access channel (MAC). Specifically, for the point-to-point case, we consider a three-party communication system, where a transmitter aims to simultaneously convey semantic information to an information receiver and energy to an energy harvesting receiver (ER). An achievable and a converse region in terms of information and energy rates are presented for both the discrete memoryless (DM) and Gaussian channel. For the DM channel, the achievable region is obtained by utilizing the asymptotic equipartition property and a converse region is obtained by using outer bounds on the semantic information rates. For the Gaussian channel, we characterize an achievable region by applying a power splitting technique between the information and the semantic context parts. A converse region is obtained that provides an estimate on the information-energy capacity while taking into account semantics. On the other hand, for the Gaussian MAC case, we consider an hybrid setup where a semantic transmitter and a conventional transmitter are employed subject to an energy harvesting constraint at the ER. Specifically, we characterize the semantic-bit information energy region, by providing an achievable and a converse region. Numerical results show that in both cases a higher performance can be achieved in terms of information and energy rates when considering a low semantic ambiguity code in comparison to the classical coding scheme (without semantic). Moreover, in the context of Gaussian MAC, it is shown that it is preferable to use semantic communications in scenarios with low signal-to-noise ratio (SNR), while conventional communications is more suitable at high SNRs.