Enhancing High-Speed Train Maintenance through Augmented Reality and Deep Reinforcement Learning: A Hybrid Optimization Approach

Abstract

High-speed rail systems demand intelligent maintenance strategies that effectively integrate real-time data analytics with human expertise. While conventional reinforcement learning (RL) models rely exclusively on IoT data, and augmented reality (AR) solutions typically operate in isolation from adaptive artificial intelligence (AI), this study proposes a novel hybrid framework that combines IoT-driven analytics, Deep Q-Learning, and immersive AR interfaces (Magic Leap 2 and Vuforia). This integration enables technician-in-the-loop optimization of maintenance workflows through a closed-loop learning process that incorporates real-time feedback. The framework was validated over a three-month period under simulated and semi-operational conditions representative of a Moroccan high-speed rail environment. Experimental results demonstrate significant improvements in key performance metrics: a 34-47% reduction in intervention time, a 55-70% decrease in human error rates, and a 28-42% decline in failure-related costs. Compared to traditional RL-only or AR-only approaches, the proposed system achieved 15-20% better efficiency gains, highlighting its advantage in real-world maintenance scenarios. These findings demonstrate the potential of AR-integrated AI to improve critical infrastructure maintenance.