Adaptive production strategy in vertical farm digital twins with Q-learning algorithms
Yujia Luo, Peter Ball
Abstract
Urban food production can contribute to sustainable development goals by reducing land use and shortening transportation distances. Despite its advantages, the implementation of digital twin (DT) technology for urban food systems has received less investigation compared to manufacturing. This article examines the influence of DT technology on adaptive decision-making in urban food production, focusing on the "Grow It York" case study. Utilising mixed integer linear programming (MILP) and Q-learning models, this study explores how DT data enhances production decisions regarding service level and resource utilisation under demand fluctuations. The findings highlight that the Q-learning model achieves up to [Formula: see text] demand fulfillment compared to [Formula: see text] for the MILP model, demonstrating a significant improvement in operational efficiency. Additionally, electricity usage per fulfilled demand is reduced by approximately [Formula: see text], advocating for broader DT application to the synergy between economic resilience and environmental sustainability. Future research directions include scaling DT implementation to manage complex supply chains, including advancing real-time data integration and incorporating sustainability considerations at supply chain level.