Agricultural Autonomous Decision-Making System "Fuxi Brain" Based on Generative Large Model Fusion Internet of Things

Abstract

Precision decision-making in agricultural production has always been a key bottleneck issue constraining the development of modern agriculture. The traditional decision-making model, which relies on manual experience, has become insufficient to meet the intelligent demands of modern agriculture. This research integrates agricultural IoT with generative large models to realize a fully autonomous decision-making agricultural intelligence system called the 'Fuxi Brain.' It consists of two parts: the first establishes a comprehensive 'Sky-Space-Earth-People-Machine' data collection system, achieving digital perception of all elements in agricultural production. The second part focuses on the intelligent decision-making system. First, within the dynamic decision-making layer of the brain, a multi-agent collaborative architecture based on multiple models (general large models and specialized agricultural models) is proposed. A Dynamic Optimal Matrix Algorithm (DOMA) is also designed to significantly enhance the system's decision-making efficiency. Finally, a full-modal alignment training method is developed to effectively address the challenges of multi-source heterogeneous data fusion. Experimental results show that in standard evaluations such as AlpacaEva and MT-Bench, the decision-making accuracy of this system improved by 36.7 percentage points compared to mainstream models like ChatGLM. The full-modal alignment training method outperformed traditional methods in cross-modal understanding tasks. The test results of the one-stop agricultural service decision-making platform show that the accuracy rate compared with the results of manual experts is 92.3%. In a practical application at the 1,367-acre corn planting area of the Dahewan Farm in Inner Mongolia, the system autonomously generated 127 decisions throughout the complete production cycle, achieving an accuracy rate of 89.7% and successfully realizing accurate autonomous decision-making from planting to harvesting. This study provides an innovative technological path and practical example for developing agricultural intelligence.

Keywords: Generative large models, Internet of Things in agriculture, Autonomous decision-making in agriculture, High precision, Food productiont