Diffusion Model Based Resource Allocation Strategy in Ultra-Reliable Wireless Networked Control Systems

TL;DR

Addresses the joint optimization of blocklength $m_i$, sampling period $h_i$, and packet error probability $p_i$ in Wireless Networked Control Systems under URLLC and finite-blocklength constraints. Proposes a two-stage diffusion-based approach: first collect an optimization-theory-based dataset, then train a conditional DDPM to map CSI to blocklength decisions, enabling inference from a Gaussian prior via a denoising process conditioned on CSI. The method achieves near-optimal total power consumption and up to an $18\times$ reduction in constraint violations compared with DRL baselines, with scalable execution times and strong resilience to policy violations. This work supports practical deployment for 6G-era URLLC in WNCS and suggests avenues for online diffusion-based adaptation in larger, more complex wireless settings.

Abstract

Diffusion models are vastly used in generative AI, leveraging their capability to capture complex data distributions. However, their potential remains largely unexplored in the field of resource allocation in wireless networks. This paper introduces a novel diffusion model-based resource allocation strategy for Wireless Networked Control Systems (WNCSs) with the objective of minimizing total power consumption through the optimization of the sampling period in the control system, and blocklength and packet error probability in the finite blocklength regime of the communication system. The problem is first reduced to the optimization of blocklength only based on the derivation of the optimality conditions. Then, the optimization theory solution collects a dataset of channel gains and corresponding optimal blocklengths. Finally, the Denoising Diffusion Probabilistic Model (DDPM) uses this collected dataset to train the resource allocation algorithm that generates optimal blocklength values conditioned on the channel state information (CSI). Via extensive simulations, the proposed approach is shown to outperform previously proposed Deep Reinforcement Learning (DRL) based approaches with close to optimal performance regarding total power consumption. Moreover, an improvement of up to eighteen-fold in the reduction of critical constraint violations is observed, further underscoring the accuracy of the solution.

Figures (4)

• Figure 1: The DDPM-based resource allocation algorithm.
• Figure 2: a) Q-Q plot for the true and generated samples. b) Testing results for different algorithms.
• Figure 3: a) Average power consumption in the testing phase. b) Average execution time for different algorithms.
• Figure 4: Average number of violations for different algorithms as a function of the number of nodes.