VMT DeStack

An active stereo sensor above the station projects patterns into the measuring volume. Its own illumination makes the measurement independent of ambient light, and exposure-time modulation captures black to white containers with one single setting. Dense point cloud at 3 m working distance.

The point cloud is reduced to the top layer and converted to a grey-scale image for faster processing. A defined search direction sets the depalletizing order and identifies the best container to grip. AI then determines the container type from inner surface, side lengths and hole pattern.

Type and position are validated against the hole spacing, with a bundle adjustment across the individual features. Three of four holes are enough, and container distortion is tolerated. Output: X, Y, Z, RX, RY, RZ. Each processing stage has its own fallback level, covering partially occluded containers, unconfirmed types, excessively tilted containers and objects detected inside the gripping holes.

The application runs on the VMT MSS software platform and can be extended with project-specific checks without touching the core process.

Search direction and picking order are freely defined, front to back or top to bottom, matching the downstream conveyor and gripper.

Lid type and position, container fill level, empty containers, empty pallets, and label position via a separate 2D camera for correct orientation on the conveyor.

A single sensor above the station covers the full depalletizing area. No sensor on the robot, no in-cycle recalibration, no second measuring step.

Gripper developed for load carriers, gripping vertically into the container holes, up to 25 kg. Sensors monitor finger travel and confirm engagement, so tilted containers or objects inside a gripping hole are detected before the gripper is damaged.

Windows IPC with a direct sensor link, fieldbus communication straight to the robot, technology packages available for KUKA, ABB and Fanuc.