Robotic bin picking is a key technology in industrial automation, which cannot always be implemented reliably. The diverse challenges include sorted parts, partially organized layers or completely unsorted bulk items, often in combination with different packaging materials such as plastic films or interlayers made of a wide variety of materials.

Fraunhofer IFF started the research project Adaptive Robotics for Robust Bin Picking and Precise Machine Tending ADAPT to address exactly these challenges. “The decisive factor for cost-effective automation is the complete system’s reliability. Many tasks that seem easy to people are tremendously challenging for robots,” explains Magnus Hanses, project manager at Fraunhofer IFF.

Several hurdles must be overcome to automate bin picking processes. One major challenge is reliable handling of packaging materials. Bins are frequently lined with plastic films that surround all parts to protect them from soiling or corrosion. Such films must be detected, grasped and securely pulled over the bin rim before the actual picking process can begin. What is more, paper or cardboard interlayers and spacers inserted between the individual layers of parts have to be automatically detected and removed. All these steps have normally required human action or have only been automatable with very high integration costs.

Another hurdle is the complete emptying of bins. Current systems are frequently incapable of automated picking of all parts, especially when the bin is largely empty and the remaining parts are in inconvenient locations or hard to access. This results in still having to keep staff on hand for final emptying despite automation. “Only when the complete process functions without human action can automation reach its full economic potential,” Hanses stresses.

What is more, parts not only have to be picked singly but also in the correct direction to be able to place them precisely afterward, whether that be in a machine tool, an assembly cell or another process step. The cycle time and accuracy required vary considerably from use case to use case.

The ADAPT project aims to address these challenges by using state-of-the-art AI methods, focusing on:

• development of robust image processing algorithms that function reliably even under difficult conditions, for instance, when surfaces are reflective, lighting is variable or there are partial visual obstructions,
• research of adaptive handling strategies that can respond flexibly to different scenarios, from handling packaging materials to manipulating hard-to-reach parts by emptying bins optimally and
• easy integration of the technologies developed in existing manufacturing environments to ensure broad applicability while minimizing any modifications required.

One particular concern is the cost effectiveness of the solutions developed. “The conversion costs for automating bin picking processes are often underestimated. A system that can flexibly handle different part types without requiring complex reprogramming provides decisive economic benefits,” explains Hanses.

The ADAPT project deliverables are particularly intended to benefit midsize companies that have shied away from extensive automation because of the high integration costs and existing solutions’ limited flexibility. The modular design of the solution being pursued and the focus on easy integration in existing systems will permit flexible modification for different use scenarios in manufacturing.