In robotic piece picking, you may have come across the word entitlement. Entitlement describes the degree to which the robot has been entitled or granted the right to handle a specific product or SKU (abbreviation) in the warehouse. Some may also use blacklists or whitelists, which is a set of items that the robot is allowed to handle or not.
What determines which products fall into either of these two lists is dependent on the robot’s ability to prove that it has a sufficiently high probability to successfully pick and handle those products, either by demonstrating its capabilities on a set of representable objects.
Table of contents
The ideal degree of entitlement is really a straightforward matter, more entitlement is always better. The more items you can identify and find good pick pose candidates, the better the efficiency of your picking cell. In the realm of piece-picking, the range of items that can be encountered varies significantly, but it is practically always in terms of thousands of SKUs (stock-keeping units), and it is not uncommon for logistics warehousing to hold millions of unique items in inventory.
In the world of manufacturing, the range of possible items the robot will encounter is much more constrained. Additionally, many of these industrial items, such as screws, billets, casings, cogs, etc., will have an associated CAD model for the picking cell to reference for detection. This is not the case for most piece-picking applications. Perfume companies do not offer CAD models for their perfume boxes; neither do makers of apparel, groceries, or toiletry items. You get the picture. In some cases, you will also receive deformable items in soft plastic bags; in that case, a picture will not be helpful either. Things arrive, and your picking cell must quickly make sense of what it is.
Typical robotic piece-picking cell in a fulfillment center.
The entitlement of your cell depends on the choice of vision system:
Without high-quality machine vision, you will face a range of challenges during deployment in a warehouse facility. The limitations of the vision system will impact the type of items that can be reliably handled with the picking cell. A low-end 3D camera will not be able to handle transparent items or items wrapped in transparent materials such as polybags or bubble wrap. This has significant consequences as most items in a warehouse inventory are typically covered in some form of transparent material. Logistics facilities that use low-end 3D tend to limit their picking use for simple cardboard boxes and similar items. These cells can then only pick a fraction of the inventory in the warehouse, resulting in an overall entitlement of perhaps 10%. This results in a small amount of inventory being routed to these automated picking cells and leaving the majority to be serviced by human pickers.
Until recent machine vision advances, picking cells were mostly staffed by human pickers.
Machine vision vendors with a clear focus on innovation, like Zivid, have taken both 3D and 2D machine vision to the next level. These new innovations go into territory that was previously considered impossible until recently. A 3D machine vision camera, such as the Zivid 2+, can now reliably capture complete point clouds with high accuracy, low noise, and 3D data without reflections and artifacts. These cameras are ushering in an era where one camera family can be used for practically any robotic manipulation task in both consumer goods and the manufacturing industries. With the very best cameras also offering coherent, high-fidelity 2D images in addition to 3D, these cameras have almost no restriction on the items that can be imaged successfully.
Warehousing and logistics are high-speed activities, with tens of thousands of items being handled and transported each day. When errors occur, they impact productivity and can cause significant headaches during troubleshooting. The kinds of errors that can occur are:
All these errors are costly in terms of actual material cost, lost time and efficiency, and maybe worst of all, customer brand sentiment if they receive damaged goods or missing goods. As we have explained, high-quality 3D cameras go a long way to solving this problem. However, an often-overlooked feature of even a high-quality camera is its level of stability in changing environments. Lighting levels and ambient temperatures can change, and vibration can occur. These factors will impact the performance of even the best spec’d 3D cameras.
These high-precision instruments need to be designed to be resilient to such environmental changes. What is needed is a camera that can deliver superb point cloud completeness and correctness, and it must be able to keep doing that regardless of variability in its environment. Robotic specialists, automation specialists, and their end customers want machine vision that can be set up once for best performance and then deployed without concern for lighting levels, new SKU items, and fine-tuning from location to location.
Logistics and fulfillment facilities are significant investments for their owners. Making a few thousand dollars in savings on machine vision is a dangerous gamble. The cost of unreliable operation will quickly surpass the upfront cost of investing in high-quality, reliable industrial 3D vision. When it comes time for the inevitable factory acceptance test (FAT), there lies the risk that the low-end 3D cameras will lead to a failure of the FAT, and automation experts know this. What has been really limiting the adoption of 3D machine vision at scale is the technology just was not there. The introduction of products such as Zivid 2+ that can consistently capture complete high-quality data on anything, including transparent objects, means the automation landscape is changing.
Logistics and e-commerce fulfillment companies are now seeing the impact of truly seeing everything and picking across the full range of SKUs in their inventories. The days of slightly lonesome, sorely limited robotic picking cells tucked away in the corner working on just cardboard boxes all day are now being relegated to the past.