University of Cambridge engineers have developed the Vegebot, a robot that uses machine learning to harvest iceberg lettuce. The design is particularly notable because, until now, iceberg lettuce has been a riddle that automation has been unable to solve.
The robot was initially developed as a proof-of-concept to harvest lettuce in a lab, but according to a new report in The Journal of Field Robotics, the Vegebot has successfully plucked the tricky vegetable in a variety of field condition.
The Vegebot identifies the lettuce head using a computer vision system that includes an overhead camera, which takes a picture of the vegetable and assesses whether or not it is ready to harvest. It also determines quality; for example, is the head of lettuce diseased? A second camera helps guide a blade mounted to a UR10 robot arm from Universal Robots. The end effector (gripper) is also used to pick it up off of the ground.
While the test was a success, the researchers admit that the prototype remains slower and less efficient than human workers.
Automated iceberg lettuce harvesting is challenging because it is damaged easily and grows flat to the ground.
The researchers employed machine-learning algorithms to teach Vegebot how to identify diseases using a series of images. The team also showed the robot how to grip the crop without dropping or damaging it.
According to the researchers, robotic harvesters like the Vegebot will not only help with labor shortages but also reduce food waste – future iterations will be able to continuously identify and harvest ripe vegetables, rather than pick the entire field in a single pass.
