Resource Allocation for Semantic-Aware Wireless Communication Networks with Imperfect CSI

In this paper, we introduce a novel uplink semanticaware wireless communication system, catering to multiple users by leveraging a shared probability graph between an access point (AP) and a base station (BS). Under imperfect channel state information (CSI), all users transmit data information to the AP through conventional bit transmission and the lower bound of signal estimation error is derived. An edge server equipped with the AP and a cloud server equipped with the BS execute information compression and recovery based on the shared probability graph, respectively. While semantic information compression incurs computational resource consumption, it significantly reduces communication resource usage. This paper addresses the challenge of minimizing system latency through jointly optimizing communication resource allocation, channel truncation threshold, and data compression ratio, considering limited wireless resources, signal estimation error, and the system’s energy budget. To solve the formulated non-convex time minimization problem, we decompose the optimization problem into four subproblems and solve each of them iteratively. In particular, the power allocation subproblem is transformed into a convex time allocation optimization problem. The bandwidth allocation subproblem is proven to be convex. The optimal channel truncation threshold is obtained through a bisection search. The data compression ratio optimization subproblem is a non-convex integer programming problem solved by the Gurobi optimizer. Numerical results show the effectiveness of the proposed algorithm, validating the necessity to consider imperfect CSI and semantic transmission and the superior performance of semantic communication compared to conventional bit transmission.