Abstract
Abstract This paper presents a physics-informed simulation study on improving Distributed Antenna Systems (DAS) performance in underground mining environments using Artificial Intelligence (AI). The simulation models dust, humidity, and metallic structures, which were found to degrade communication by reducing signal stability by 15–20%, adding up to 12 dB of attenuation at humidity above 70%, and increasing interference by 30–35%. An AI-based adaptive control pipeline was implemented, combining multilevel learning for power and tilt control, meta-learning for rapid adaptation to changing conditions, and adaptive beamforming for interference suppression. In mining-like conditions, this pipeline increased effective coverage from 72.0% to 90.0%, reduced adaptation time from 2.00 s to 1.64 s, and lowered packet loss from 10.0% to 8.0%. The models achieved a mean squared error below 0.03 and prediction accuracy up to 92%, indicating that AI-driven DAS control can improve underground communication reliability by 20–25% and has potential to reduce production-related costs by about 30%.