Active Vibro-Tactile Sensing for Non-Destructive Selective Harvesting of Strawberries in Precision Agriculture

Abstract

In recent decades, precision agriculture has attracted growing interest, both for reducing farmers’ manual workload and for supporting the Sustainable Development Goal of achieving zero hunger. With continued population growth, the development of novel technology-driven harvesting methods has become increasingly important. One promising direction is the integration of intelligent robotic systems into orchards and greenhouses to monitor crop conditions and perform harvesting in a timely manner. To achieve this, robots must be equipped with robust sensory feedback. For ripeness estimation, vision-based methods are the most widely used; however, they often suffer from occlusions caused by branches and foliage and may provide unreliable assessments for fruits whose ripeness is not clearly visible from external appearance alone. In contrast, touch offers a natural means of assessing fruit firmness, which is closely related to ripeness. Traditional tactile approaches, however, often rely on passive procedures such as squeezing and observing indentation, which can bruise soft ripe fruit. This thesis proposes a non-destructive tactile sensing method based on active vibrational feedback, which enables a robot to establish light contact with a strawberry and estimate its ripeness prior to grasping. The sensor was further integrated with a robotic arm and gripper to emulate a realistic selective strawberry-harvesting scenario with adaptive grasping based on the estimated ripeness. Through vibrational signal analysis and statistical validation, the system achieved a classification accuracy of 71\% on previously unseen strawberry samples. These results demonstrate the potential of the proposed approach for affordable and scalable robotic harvesting.