AI-Driven Radio Propagation Prediction in Automated Warehouses Using Variational Autoencoders

The pervasive demand for data-intensive applications and the rapid integration of emerging technologies are driving an unprecedented transformation in wireless communication, particularly within Industry 4.0. Optimizing 5G and future networks for automated environments like smart warehouses requires advanced solutions for indoor radio propagation. To this end, this paper introduces WISVA (Wireless Infrastructure for Smart Warehouses using VAE), an AI-based framework utilizing a novel Variational Autoencoder (VAE) model (AI Contribution). The VAE's unique architecture learns complex electromagnetic (EM) wave interactions from meticulously crafted, physics-informed data tensors, enabling it to accurately model signal behavior impacted by diverse obstacles. This engineering application provides precise signal-to-interference-plus-noise ratio (SINR) heatmaps for 5G wireless bands in automated Industry 4.0 settings. We demonstrate WISVA's remarkable robustness and adaptability through its superior performance in denoising tasks, validation, and critically, extrapolation to entirely unseen warehouse layouts and configurations. Comparative analysis via reconstruction error heatmaps reveals WISVA's significantly higher accuracy against traditional autoencoders, establishing its potential as a critical enabler for efficient wireless infrastructure optimization in Industry 4.0.