An Intelligent LLM–Fuzzy Decision Support and Digital Twin Framework for Adaptive Agricultural Robotics Systems

This study develops an integrated LLM–fuzzy decision-support and digital twin framework for adaptive agricultural robotics, combining semantic reasoning, fuzzy inference, multi-objective optimization, and feedback-driven state updating. Numerical sensor data, operational information, and textual field knowledge are transformed into validated decision rules and model-ready parameters, enabling coordinated decisions across economic, operational, crop-stress, and sustainability-related objectives. The framework is evaluated as a simulation-based proof of concept calibrated with real observations from a semi-smart tomato farm, while controlled digital twin scenarios are used to assess its behavior under environmental and operational disturbances. Compared with numerical optimization alone, the complete LLM–fuzzy framework reduces operational cost by 3.35% and crop stress by 7.98%, while improving operational performance by 4.45%. Ablation, repeated-run, scenario, sensitivity, and dynamic-adaptation analyses further demonstrate consistent computational performance under the evaluated conditions. Sustainability is assessed through a model-based measure integrating water use, energy consumption, and soil-compaction-related impacts. These findings demonstrate the potential of combining semantic reasoning, uncertainty processing, optimization, and digital-twin feedback for adaptive agricultural decision support. However, the results represent a calibrated simulation-based proof of concept rather than validation through a full-scale autonomous field deployment; broader multi-season and field-scale validation is therefore required before generalizing the reported operational and sustainability benefits.