In short
Mixed pallets arrive with different products, box sizes, orientations and stacking patterns. Today, unloading them requires manual handling, because the position and orientation of each item can't be predicted in advance.
The goal is to automate depalletizing with a robot combined with 3D vision and flexible gripping: identify accessible objects on the pallet, determine a suitable grasp, pick them reliably, and place them on the downstream conveyor or a defined destination — without requiring the pallet configuration to be known beforehand.
How the process works
- 1A pallet containing mixed products arrives at the depalletizing station.
- 2Detect — a 3D vision system captures the current pallet state and identifies objects that can be picked.
- 3Select the next object — the system determines which object is accessible and calculates a suitable grasp pose.
- 4Pick — the robot approaches the object and picks it with a suitable gripper, for example vacuum or another flexible gripping technology.
- 5Transfer — the robot moves the product from the pallet to the defined drop-off position.
- 6Verify — the system confirms the object was removed successfully and updates the pallet state.
- 7Repeat — the next accessible object is selected and the process continues automatically, until the pallet is empty.
Requirements
Must have · 7
Mixed pallets
Handle pallets without requiring a predefined stacking pattern.
Object detection
Detect individual products and estimate their position and orientation from sensor data.
Flexible gripping
Handle different product dimensions, orientations and surfaces with as few tool changes as possible.
Collision-free motion
Generate robot motions that avoid the remaining products, the pallet and surrounding equipment.
Pick verification
Detect failed or incomplete picks and recover without unnecessary operator intervention.
Unknown pallet state
Determine the current pallet configuration automatically rather than relying exclusively on upstream pallet data.
Industrial integration
Provide interfaces to the robot, the safety system and the downstream material flow.
Should have · 3
Automatic recovery
Retry failed picks using an alternative grasp or approach strategy.
New product handling
Allow new product variants to be introduced with limited manual configuration.
Process monitoring
Record picks, failures, cycle times and other relevant system states.
Main challenges
Challenges
5 open. Each one can be answered on its own — a proposal does not have to solve the whole cell.
- 1
Reliable perception
Products may have different colors, textures, geometries and orientations. Objects can partially occlude each other, which makes segmentation and pose estimation harder as the pallet changes.
- 2
Finding a grasp
Detecting an object is not enough — the system must identify a surface the gripper can actually reach and use reliably. For vacuum gripping, that means finding sufficiently large, accessible and suitable surfaces for the suction cups.
- 3
Choosing the next object
Not every detected object should be picked immediately. The system needs to determine a feasible picking sequence based on accessibility, grasp quality and collision risk.
- 4
Handling product variation
The solution should avoid requiring extensive engineering whenever a new product variant is introduced. A key question is how much prior information about each SKU is actually required.
- 5
Recovering from failed picks
Products may move, slip, deform or fail to attach to the gripper. The system should detect these situations and decide whether to retry, select another grasp, or request operator assistance.
Decisions still open
The final system concept depends on the product range and the operating environment:
- 3D camera technology and camera position
- Fixed camera vs. a camera mounted on the robot
- Vacuum gripper vs. an adaptive gripping concept
- SKU database vs. model-free object detection
- Grasp-generation approach
- Robot payload and reach
- Required cycle time
- Downstream placement strategy
- Handling of damaged or difficult products
What a useful proposal covers
A useful solution concept explains how the system would handle:
- Detection and segmentation of products on an unknown mixed pallet
- Grasp generation and selection
- Flexible gripping across the expected product range
- Collision-free robot motion
- Failed-pick detection and recovery
- Integration with the downstream material flow
- Introduction of new product variants
- Expected cycle time and system limitations
It should also state plainly what information about the products must be known beforehand, and what the system is expected to work out on its own. A proposal can address the complete depalletizing cell, or focus on a single challenge — perception, grasp planning, gripping, or motion planning.
Submit a solution
Have an idea how to solve this, or a part of it? Send it to Birdwave: a comment, and a document if you have one — a concept, a layout, a datasheet. Only Birdwave sees what you send: not the company, not other members.
Sign in to submit a solution.
Sign inQuestions & comments
Sign in to ask a question about this use case, or to read and join the discussion.
Sign in