CNN-Enabled Scheduling for Probabilistic Real-Time Guarantees in Industrial URLLC

Efficient resource allocation in multi-cell, multichannel wireless networks is essential for supporting real-time applications with strict latency and reliability requirements. In this work, we enhance the Local Deadline Partition (LDP) algorithm by introducing a CNN-based dynamic priority prediction mechanism that improves interference management and scheduling decisions. Unlike conventional LDP with static priorities, our approach leverages a Convolutional Neural Network and graph coloring to adaptively assign link priorities based on real-time traffic, available transmission opportunities, and network conditions. With offline training, the method incurs minimal inference overhead while enabling more efficient resource allocation, resulting in improved network capacity, SINR, and schedulability. The simulation results demonstrate SINR gains of up to$\text{1 1 3 \%, ~} \text{9 4 \%}$, and 49% in three network configurations, highlighting the effectiveness of the proposed approach for complex industrial wireless scenarios.