Fruit Picking Robot Arm Training Solution Based on Reinforcement Learning in Digital Twin

In the era of Industry 4.0, digital agriculture is developing very rapidly and has achieved considerable results. Nowadays, agriculture-based research more focused on use robotic fruit picking technology, main direction such topics algorithms for computer vision. However, when vision successfully locate target object, it still necessary to arm movement reach object at physical level, but path planning received minimal attention. Based this deficiency, we propose Unity software as a twin platform plan ML-Agent plug-in reinforcement learning means train path, improve accuracy fruit, happily effect method much improved than traditional method.