Dynamic light optimization in vertical farming using an IoT-driven digital twin framework and artificial intelligence

Abstract

The global agricultural sector faces mounting challenges from climate change, population growth, urbanization, and environmental degradation, necessitating innovative solutions to ensure food security. Urban and peri-urban agriculture, particularly vertical farming, offers a sustainable approach to increase food production while minimizing land use, reducing environmental impact, and enhancing resource efficiency. Unlike conventional vertical farming systems that rely on static spectral recipes with fixed light compositions (e.g., Red-to-Blue ratios derived from historical data), this study introduces an Internet of Things-enabled smart vertical farming system that leverages digital twin technology and a genetic algorithm (GA) to dynamically optimize lettuce growth by adjusting RGB LED spectra throughout the crop cycle. The system monitors and controls key environmental parameters within a growth tower, including temperature, humidity, and lighting. A digital twin facilitates real-time data exchange between physical and virtual components, while the GA iteratively refines the light composition. Over a 34-day cultivation period, the algorithm identified an optimal RGB configuration (R:211, G:169, B:243; maximum intensity: 255) that aligns with spectral values reported in literature for lettuce, despite not directly measuring photobiological metrics such as Photosynthetic Photon Flux Density. To our knowledge, this is the first study to implement a dynamic, GA-driven spectral optimization strategy in vertical farming. While the objective was not to surpass traditional static lighting recipes, the results validate that adaptive methods can reliably converge to established optima. The IoT platform demonstrated robust capabilities in data collection, processing, and actuation, underscoring the promise of adaptive lighting strategies for controlled agriculture. Future research will focus on incorporating additional spectra (e.g., deep red, ultraviolet), automating data collection via image recognition, and analyzing energy efficiency to enhance scalability.