Toward autonomous blackberry harvesting with a soft gripper and vision-controlled robotic arm

Robotic harvesting has become a significant topic in recent years as it addresses labor shortages, reduces production costs, and enhances food quality. In this work we present a comprehensive framework for robotic blackberry harvesting. It employs a low-cost 6-DOF robotic arm paired with a soft inflatable gripper specifically designed for blackberries and it leverages the capabilities of YOLOv8 for vision-based control. A standardized set of metrics and tests to objectively evaluate the performance of robotic harvesting in controlled conditions and to inform field deployment is introduced. Our system reached peak performances of 98.4% for the vision component and of 76.6% for grasping effectiveness, with a combined success rate of 52% for the whole pipeline, but with significant variability depending on the pose of the blackberry. The multipurpose arm showed significant limitations, suggesting the development of specialized hardware for future deployment. Our results demonstrate the system’s potential for robotic harvesting of blackberry, while suggesting areas for future research.