WISVA: Generative AI for 5G Network Optimization in Smart Warehouses
Authors:
Rahul Gulia,
Amlan Ganguly,
Andres Kwasinski,
Michael E. Kuhl,
Ehsan Rashedi,
Clark Hochgraf
Abstract:
The next decade will usher in a profound transformation of wireless communication, driven by the ever-increasing demand for data-intensive applications and the rapid adoption of emerging technologies. To fully unlock the potential of 5G and beyond, substantial advancements are required in signal processing techniques, innovative network architectures, and efficient spectrum utilization strategies.…
▽ More
The next decade will usher in a profound transformation of wireless communication, driven by the ever-increasing demand for data-intensive applications and the rapid adoption of emerging technologies. To fully unlock the potential of 5G and beyond, substantial advancements are required in signal processing techniques, innovative network architectures, and efficient spectrum utilization strategies. These advancements facilitate seamless integration of emerging technologies, driving industrial digital transformation and connectivity. This paper introduces a novel Variational Autoencoder (VAE)-based framework, Wireless Infrastructure for Smart Warehouses using VAE (WISVA), designed for accurate indoor radio propagation modeling in automated Industry 4.0 environments such as warehouses and factory floors operating within 5G wireless bands. The research delves into the meticulous creation of training data tensors, capturing complex electromagnetic (EM) wave behaviors influenced by diverse obstacles, and outlines the architecture and training methodology of the proposed VAE model. The model's robustness and adaptability are showcased through its ability to predict signal-to-interference-plus-noise ratio (SINR) heatmaps across various scenarios, including denoising tasks, validation datasets, extrapolation to unseen configurations, and previously unencountered warehouse layouts. Compelling reconstruction error heatmaps are presented, highlighting the superior accuracy of WISVA compared to traditional autoencoder models. The paper also analyzes the model's performance in handling complex smart warehouse environments, demonstrating its potential as a key enabler for optimizing wireless infrastructure in Industry 4.0.
△ Less
Submitted 16 June, 2025;
originally announced June 2025.
