Mixed-Case Depalletizing with a Robot and 3D Vision: How It Works

A pallet that never arrives the same way twice

In a distribution warehouse or at a factory's raw materials receiving area, pallets arrive from different suppliers, with boxes of various sizes, stacked without a fixed pattern and often with tape, labels or film. Someone has to unload them box by box onto a conveyor or a picking station. It is one of the heaviest and most repetitive jobs in logistics.

For years, automatic depalletizing was only viable when pallets were homogeneous: same box, same pattern, always identical. The robot repeated programmed positions. With mixed pallets, that does not work. 3D vision has changed the situation. This article explains how mixed-case depalletizing works, which conditions make it viable and which questions to ask before tackling it.

Palletizing and depalletizing are not the same problem

Palletizing starts from a controlled situation: boxes arrive on a conveyor, one at a time, with known dimensions, and the robot places them following a pattern. We explain it in the article on automatic palletizing solutions.

Mixed-case depalletizing is exactly the opposite: the robot does not know in advance what is on the pallet or where each box is. It has to find out every cycle. That is why vision is not an accessory, but the core of the solution.

How a 3D vision depalletizing system works

• Capture of the top layer. A 3D camera above the pallet generates a point cloud of the visible surface.

• Box segmentation. The software identifies each individual box, even when they sit together with no visible gap, and calculates its position, orientation and approximate dimensions.

• Box selection. It chooses the next box that can be removed without destabilizing the others, usually starting with the highest layer and the most accessible boxes.

• Pick and place. The robot picks the box, places it on the conveyor or at its destination and the cycle repeats.

The key point is that the system does not need to know the box catalogue in advance. Deep learning segmentation algorithms make it possible to recognize new boxes without programming them one by one, although it is always worth validating with the warehouse's real formats.

Which conditions make the project viable

Packaging type

Corrugated boxes with flat faces are the most favourable case. Boxes with very similar printing, boxes packed together with no gap, shrink-wrapped surfaces, bags or boxes deformed by weight are more demanding and should be tested.

Box weight and size

These determine the choice of robot. With light and medium boxes and moderate throughput, a cobot with a lifting column may be enough. With heavy boxes, tall pallets or high throughput, an industrial robot with greater payload and reach is needed.

Pallet height

The robot and camera must cover everything from the highest layer down to the last box resting on the pallet. If the robot's vertical reach is not enough, it can be extended with a lifting column, as explained in the article on lifting columns and linear axes for cobots.

Required throughput

El ciclo incluye la captura, el procesamiento y el movimiento. Capturar la siguiente imagen mientras el robot deposita la caja anterior ayuda a no penalizar el tiempo de ciclo.

The gripper: the critical point

In mixed-case depalletizing, the most common gripper is vacuum, because it adapts to boxes of different sizes without a tool change. But there are nuances:

Zoned vacuum grippers. They activate only the suction cups positioned over the box, avoiding picking two boxes at once or losing vacuum with small boxes.

Open boxes or loose lids. They may require additional holding, since vacuum on the lid does not guarantee lifting the contents.

Heavy and large boxes. It may be necessary to combine vacuum with mechanical support underneath to prevent the box from breaking when lifted.

Gripper selection is covered in detail in the article on how to choose a cobot gripper.

Limitations you need to know

• Slip sheets and film. Cardboard slip sheets between layers and wrapping film must be removed or handled. Some installations remove them manually before the cycle; others include a dedicated tool.

• Unstable pallets. If boxes have shifted during transport, the system may find awkward grip positions.

• Identical boxes with no gap. Two identical boxes side by side may look like one if there is no visible line between them. This is one of the cases most worth validating.

• Error recovery. You need to define what happens if a box falls or cannot be picked: operator alert, reject area or controlled stop.

Workstation safety

The operator has to approach to remove the empty pallet and place a new one. The usual approach is to protect those entries with laser scanners or light curtains that slow down or stop the robot when a person enters the area. With two pallet stations, the robot can keep working on one while the other is changed.

Where to start

Mixed-case depalletizing is viable today in many warehouses and factories, but its success depends on the real formats that arrive at each site. The most useful step is to gather a representative sample of pallets, including the most difficult ones, and test with them.

At Rebots we work with Mech-Mind 3D cameras and any type of robotic arm for mixed-case (de)palletizing. You can see the approach on the 3D camera (de)palletizing solution page and tell us about your case with photos of your pallets so we can assess feasibility.