Reinforcement Learning for Smart Mobile Factory Operation in Linear Infrastructure Projects

Abstract

Mobile factories promise an increased project efficiency with on-demand production and Just-in-Time delivery of prefabricated elements. However, traditional scheduling methods predominantly focus on either factory or site and neglect the factory mobility, often leading to suboptimal synchronization. To address this gap, this paper introduces a novel reinforcement learning (RL)-based model for optimizing the operational policy of mobile factories in infrastructure projects. The developed model simultaneously schedules on-site and off-site operations, effectively integrating the performance metrics at the project level. Utilizing RL, the factory's production management system continuously learns and adjusts in response to real-time project developments, ensuring optimal decision-making regarding scheduling and resource allocation.