Vision-Based Palletizing with MoveIt Task Constructor

Manual palletizing is labor-intensive and error-prone. The client needed automated product recognition and intelligent sorting with collision-free motion planning.

• Converted the palletizing algorithms into ROS 2 and MoveIt 2 code.
• Trained YOLO on the products and turned detections into robot paths.
• Added per-pallet feedback to keep errors from accumulating.

• Palletizing algorithms are translated into ROS 2 code that drives the robot.
• A YOLO model trained on the products turns each detection into a robot path for placing it on the pallet.
• Feedback after every pallet stops errors from building up.

Reliable product recognition and automated palletizing with collision-free motion. The system handles multiple product categories with intelligent sorting.

Training YOLO on a custom product dataset required careful annotation and augmentation. MoveIt Task Constructor's API is complex — building reliable pick-and-place sub-trajectories needed extensive testing in simulation before real-world deployment.