Towards Human-Robot Interaction in Agriculture Using LargeLanguage Models

Labor shortages and usability challenges limit the adoption of robotics in agriculture.This work explores how Large Language Models (LLMs) and Vision-Language Models (VLMs) can bridge this gap by enabling non-expert users to command robots using natural language.A modular system was developed to interpret instructions, execute tasks, and generate visual field reports.Evaluations in a simulated field showed that hybrid prompting strategies yielded reliable plans, while VLMs supported effective object detection and contextual reporting.This approach reduces entry barriers to robotics and promotes accessible, intelligent agricultural automation.