A field-tested robotic harvesting system for iceberg lettuce

Area of Science:

• Robotics
• Agricultural Engineering
• Computer Vision

Background:

• Robotic systems offer unique opportunities in agriculture, particularly for autonomous harvesting in unpredictable environments.
• Harvesting iceberg lettuce presents challenges due to its susceptibility to damage and difficulty in visual detection.

Purpose of the Study:

• To develop and field-test an autonomous robotic harvesting system for iceberg lettuce.
• To address the specific challenges of handling and visually identifying iceberg lettuce for robotic harvesting.

Main Methods:

• Development of a platform named Vegebot, integrating a vision system, custom end effector, and control software.
• Implementation of a vision and learning system with two integrated convolutional neural networks for lettuce classification and localization.
• Design of a damage-free end effector with force feedback for controlled harvesting and ground detection.

Main Results:

• Successful field testing of the Vegebot system.
• Demonstrated effectiveness of the vision system in localizing and classifying iceberg lettuce.
• Validated the integrated system's capability for autonomous, damage-free lettuce harvesting.

Conclusions:

• State-of-the-art vision approaches can be effectively applied to agricultural robotics.
• Mechanical systems can be engineered to leverage environmental constraints for successful agricultural automation.
• The Vegebot system demonstrates a viable solution for autonomous iceberg lettuce harvesting.