Smart bin picking: Automatically picking up unknown objects? It's possible!

Various products are ready to be picked

In many automated production processes, robots are used to pick products from bins—this is known as bin picking. Until recently, this technique was mainly useful for known objects: products with a fixed shape, known dimensions, and an available CAD model.

In such cases, the system knows exactly what to look for. The objects need to be neatly arranged and must not deviate from the input model. But in practice, things are different.

Think of return logistics, bins with mixed products, or natural variations in shape or position. In such situations, classical bin picking doesn’t perform as it should.

That’s when we talk about picking unknown objects, which requires a different approach.

What’s the Difference? 📦

With known objects, everything revolves around predictability: the system is pre-trained for one specific object type, with data on its shape and size. Anything that deviates from that is either not recognized or not picked up correctly.

With unknown objects, that predictability disappears. Instead, the system has to determine for itself what it’s seeing, whether it’s graspable, and how to safely pick it up.

That calls for vision technology, AI algorithms, and a robot that can adapt to variation.

Why This Approach Pays Off 💡

The major advantage of bin picking unknown objects is flexibility. You don’t need to reconfigure the system every time a different product ends up in the bin.

Whether objects are upright, tilted, or even partially hidden, the system identifies opportunities for grasping on its own. This makes it ideal for processes with high variability or where speed and scalability are essential.

It also saves on setup time, there’s no need to program each object in advance. And it improves continuity, as the system requires less human intervention to keep running.

Example: cosmetics 🧪

For a client in the cosmetics sector, we developed a solution that automatically picks up items like bottles, jars, and tubes—even when they’re all mixed together.

The system scans the contents of the bin, recognizes the different shapes, and selects the most graspable object each time. The robot then picks up each item gently, without causing damage.

The result: a task that used to be manual, time-consuming, and error-prone now runs quickly, efficiently, and with high quality.

One product is selected to be picked from a choice of three different tubes.

Automation with more freedom

Smart bin picking makes it possible to automate unstructured, variable situations—not just in cosmetics, but in any sector where flexibility, precision, and speed matter.

🔗 Curious to see how this works in practice? Check out our case: heliovision.be/en/cases/binpicking