An integrated tomato harvesting framework using a hybrid soft-rigid gripper with semantic segmentation and keypoint...

Area of Science:

• Agricultural Robotics
• Soft Robotics
• Mechatronics

Background:

• Efficient and gentle harvesting of tomatoes is crucial for reducing post-harvest losses and improving food quality.
• Existing robotic harvesting systems often struggle with delicate fruit handling, variable conditions, and precise manipulation.

Purpose of the Study:

• To develop and evaluate an autonomous robotic system for tomato harvesting.
• To design a hybrid robotic gripper capable of gentle yet secure fruit grasping and pedicel cutting.
• To integrate perception, control, and motion planning for a complete harvesting cycle.

Main Methods:

• A hybrid robotic gripper combining soft auxetic fingers, a rigid exoskeleton, and a latex basket was developed.
• An RGB-D camera and Detectron2 pipeline enabled semantic segmentation and keypoint localization for ripe tomato detection.
• Closed-loop grasp-force regulation using a PID controller and force-sensitive resistors, alongside PSO-based trajectory planning for a 5-DOF manipulator, were implemented.

Main Results:

• The system achieved an average cycle time of 24.34 seconds with an 80% success rate in laboratory conditions.
• Gentle grasping forces (0.20-0.50 N) were maintained, minimizing fruit bruising.
• Successful demonstration of complete picking cycles, including approach, separation, cutting, grasping, transport, and release.

Conclusions:

• The developed hybrid end-effector and integrated system represent a practical approach to autonomous tomato harvesting.
• The system demonstrates effective perception, closed-loop control, and motion planning for robotic fruit collection.
• Further research is needed to address limitations and failure modes for successful field deployment.