Effective Throughput Maximization for NOMA-Enabled URLLC Transmission in Industrial IoT Systems: A Generative AI-Based Approach

Abstract

<p>The development of B5G and 6G technologies has led to an explosive growth in device connectivity density in Industrial Internet of Things (IIoT) systems. However, the limited spectrum resources in industrial wireless networks pose significant challenges for large-scale access and communication rates, especially for factory automation applications that are sensitive to control stability and latency. In this paper, we investigate an uplink non-orthogonal multiple access (NOMA) transmission for ultra-reliable and low-latency communication (URLLC) services in IIoT systems, where sensors in NOMA clusters transmit collected data to the base station to meet the high communication rate and control stability requirements of controlled devices. The dynamic control convergence constraint is theoretically transformed into an optimal control condition in each communication round based on the decoding error probability. Additionally, we formulate an optimization problem to maximize the effective throughput of the considered system in the finite blocklength regime by jointly optimizing blocklength allocation, power allocation, and decoding error probability. To solve this mixed integer non-linear programming (MINLP) problem, we decompose it into two sub-problems and propose an efficient optimization framework based on generative AI. Specifically, we apply successive convex approximation (SCA) to solve the blocklength allocation sub-problem, and use a diffusion model to address the joint power control and decoding error probability sub-problem. Finally, extensive simulation results demonstrate the effectiveness of this approach.</p>