LLMs Autonomously Optimize Vertical Farms, Boosting Yield and Efficiency

Researchers have developed a novel closed-loop system that leverages Large Language Models (LLMs) as autonomous co-pilots for digital agriculture. This system integrates data from a 49-channel phytosensor network, which monitors various biophysical parameters of plants, with LLM-driven control over environmental actuators. The LLM interprets real-time plant physiology data and then triggers hardware adjustments to optimize microclimates, execute phenotyping protocols, or induce controlled stress scenarios. The framework moves beyond traditional data analysis to enable autonomous, AI-guided experimentation. Validation across vertical farm and single-plant setups demonstrated its ability to decipher complex physiological fluctuations. In production-scale deployments, the LLM-controlled agents optimized parameters like biomass accumulation, chlorophyll content, and energy consumption by modulating lighting. This approach significantly improved efficiency, reducing production cycles by 35% in minimal-time mode and energy consumption by 18% in energy-optimization mode. Notably, the system autonomously discovered an unforeseen strategy for dark-induced chlorophyll accumulation, leading to 67.9% energy savings. This innovation transforms LLMs into practical tools for enhancing agricultural productivity and sustainability.