Research on System Integration Development and Experiment for a Tomato Harvesting Robot

Selective harvesting of fresh-market tomatoes is a labor-intensive task and one of the most cost-intensive stages in tomato production. Therefore, “machine substitution” is being an effective way to achieve automated harvesting for urgent need. To address the harvesting requirements of tomatoes grown on overhead suspension systems in greenhouse facilities, an automated tomato harvesting robot was developed by integrating RGB-D cameras, LiDAR sensor, a four-wheel-drive and four-wheel-steering(4WID4WIS) mobile chassis, a 6DOF collaborative robotic arm, and an integrated picking-recovery mechanism. A software system was developed based on the ROS framework to enable an autonomous “move-observe-harvest” workflow. Tomato detection and localization were implemented by using the YOLOv5 algorithm framework, while motion trajectory control of the robotic arm was achieved through a combination of hand-eye calibration, the RRT-Connect algorithm, and key waypoint strategies. Field tests of the harvesting robot were conducted in a greenhouse production environment. The results showed that the average harvesting cycle was 12.48 seconds per tomato, the harvesting success rate was 84%. These findings provide valuable insights for the development and industrial application of intelligent harvesting equipment tailored to tomatoes cultivated on overhead suspension systems in greenhouse facilities.