Rate-conforming Sub-band Allocation for In-factory Subnetworks: A Deep Neural Network Approach

This paper focuses on the critical challenge of sub-band allocation for dense 6G In-factory subnetworks. We introduce a deep learning (DL) framework explicitly designed to effectively address the inherent optimization problem in sub-band assignment to subnetworks. To enhance the model’s training process, a novel strategy is implemented to handle integer optimization variables. The proposed approach aims at utilizing resources more efficiently by maximizing the number of rate-conforming subnetworks, serving as the key component of the loss function. Simulation results demonstrate that, across various classes of subnetworks, the proposed method achieves superior performance compared to State-of-the-Art (SoA) benchmarks with minimal computation time.