Raspberry-Picking Robot Could Help Reduce Harvesting Costs

A robot that can pick raspberries without crushing or damaging the fruit — after being trained only in a lab with an artificial raspberry plant — is presented in a study published in Communications Engineering. The authors suggest that, with further development, the robot could be used to harvest raspberries throughout the day and could help to reduce harvesting costs.

The robot, developed by Kai Junge and colleagues, consists of a remote-operated wheeled base and a robotic arm, which contains two cameras and a gripper with silicone fingers and pressure sensors. The cameras, along with artificial intelligence, are used to automatically detect and estimate the size and position of raspberries according to their colour. These estimates determine the movements of the robotic arm to align the gripper with a raspberry, so it can then grip and pull the fruit off a plant. The forces applied by the robot are based on measurements the authors made of the gripping and pulling forces applied by humans when picking an artificial silicone raspberry containing pressure sensors from an artificial plant.

The robot was trained to detect and pick the artificial raspberry in a laboratory before its ability to harvest real raspberries was tested in the field. As the amount of force needed to pick a raspberry depends on its ripeness, the authors trained the robot to pick raspberries at different stages of ripeness by adjusting the attachment between the artificial raspberry and the artificial plant.

The authors found that during 25 attempts at picking real raspberries in the field without crushing them, the robot performed with an 80% success rate, with 56% of raspberries picked without sustaining damage and 24% sustaining only minimal damage. They report from conversations with raspberry growers that human fruit pickers have around a 90% success rate when picking raspberries with no or minimal damage to fruit. When the combined raspberry detecting and picking abilities of the robot were tested in the field, the authors found that the robot successfully detected and picked four raspberries within seven attempts.

The authors suggest that future research is needed to train the robot to pick individual raspberries from bunches as well as raspberries with a range of properties under a variety of different lighting, temperature, and humidity conditions.

Reference: Junge K, Pires C, Hughes J. Lab2Field transfer of a robotic raspberry harvester enabled by a soft sensorized physical twin. Communications Engineering. 2023;2(1):40. doi: 10.1038/s44172-023-00089-w

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source.