Perception-Guided Trajectory Generation and Force-Controlled Grasping for Robotic Blackberry Harvesting

Manuel Navarro-Gutiérrez (Dr) 

Tecnológico de Monterrey, Mexico

 

This talk discusses the development of a perception-guided manipulation system for blackberries growing in Mexican greenhouses. The system is capable of detecting ripe blackberries, generating smooth approach trajectories, and achieving safe grasping during harvesting. A fine-tuned YOLO detector is employed to identify fruits and estimate ripeness under varying conditions. Density-based clustering organizes detections to support trajectory planning, while pose estimation provides geometric information for manipulation. From the robot’s current configuration and a target defined by its position and associated radius, a geometrically well-conditioned approach objective is formulated. A continuous three-dimensional trajectory is generated using time-based sigmoid functions to ensure smooth transitions, and trajectory tracking is performed through a kinematic control scheme. Grasping is achieved through an indirect force control strategy implemented on a force-sensing gripper to reduce the risk of fruit damage. Initial experiments demonstrate reliable fruit detection under partial occlusion, smooth approach motions with stable trajectory tracking, and consistent closed-loop force regulation during contact.

 

Manuel Navarro-Gutiérrez received the M.Sc. and Ph.D. degree in electrical engineering in 2013 and 2018 from the CINVESTAV-Guadalajara (Mexico). His research is related to mathematical modeling of discrete event systems, continuous systems, and hybrid systems, as well as the development of controllers and their implementation for changing the qualitative and quantitative properties of these systems. He is a member of the National System of Researchers, with the Level 1 distinction.