Bit-Aware Semantic Resource Allocation

This chapter introduces an adaptive semantic resource allocation framework incorporating semantic-bit quantization (SBQ), which maintains compatibility with existing wireless communication systems while addressing the inaccuracies in environmental perception caused by the supplementary mapping association linking semantic metrics and transmission metrics. Specifically, SBQ constitutes a hybrid uniform-non-uniform quantization approach designed to enable efficient coding between semantic information and bits. The performance of SemCom networks is assessed through the novel introduction of the quality of service for SemCom (SC-QoS), a metric that includes semantic quantization efficiency (SQE) and transmission latency. A joint optimization problem is formulated to maximize the aggregate effective SC-QoS by adjusting the base station’s transmit beamforming, the bit allocation for semantic representation, together with subchannel assignment and bandwidth resource allocation. To address this non-convex problem, a hybrid deep reinforcement learning (DRL) algorithm is employed to create an intelligent resource allocation strategy, which allows the agent to adapt to dynamic wireless environments and various semantic tasks. Results from simulations indicate that the introduced methodology effectively mitigates semantic noise and attains superior wireless communication performance compared to several baseline approaches. Furthermore, in comparison with resource allocation schemes based on the mapping-guided paradigm, the proposed adaptive approach can attain a performance enhancement of up to \({13\%}\) in terms of SC-QoS.