Robust Information Bottleneck for Task-Oriented Communication With Digital Modulation

Task-oriented communications, mostly using learning-based joint source-channel coding (JSCC), aim to design a communication-efficient edge inference system by transmitting task-relevant information the receiver. However, only without introducing any redundancy may cause robustness issues in learning due channel variations, and JSCC which directly maps source data into continuous input symbols poses compatibility on existing digital communication systems. In this paper, we address these two first investigating inherent tradeoff between informativeness of encoded representations distortion received representations, then propose task-oriented scheme with modulation, named discrete (DT-JSCC), where transmitter encodes features representation transmits it receiver modulation scheme. DT-JSCC scheme, develop robust encoding framework, bottleneck (RIB), improve derive tractable variational upper bound RIB objective function approximation overcome computational intractability mutual information. The experimental results demonstrate that proposed achieves better performance than baseline methods low latency, exhibits variations applied framework.