The AI-supported optical systems for quality assurance from ivii open up new possibilities for supporting people in their working environment. Peter Stelzer, CEO of ivii, explains how "cognitive assistance" works and in which industries and areas it is already being used successfully.
Your company is known for innovative optical systems for quality assurance. The motto is: "Bridging the Reality Gap". What does that mean?
We are used to optimizing processes based on key figures that come from machine or control data. However, important key figures are missing here - namely those that are directly related to reality: What is really in the container? Does the content correspond exactly to what it should be according to the higher-level software or planning system (WMS, WCS, MES, PPS, etc.)? Or: Is the quality of a produced item really as good as it says on the docket? With our systems, it is possible to record "reality key figures" and enrich the existing key figures with them. This closes the "reality gap" and allows processes to be better optimized.
Optical inspection is your main focus. What advantages does it offer over other sensors or systems?
There are many sensors for data acquisition, such as scanners, scales etc.. However, these only ever record one specific aspect. An optical sensor, on the other hand, is a holistic source of information: I can not only detect the dimensions and weight of an item, but also whether the items in a container are intact, or whether the right products are in the container at all - which is becoming increasingly important. Because the direct distribution channel - often from the producer directly to the end customer - means that there is no intermediate trade and therefore no control authority. These checks must therefore be integrated directly into the supply chain - and we can do this with our systems.
Do you use special cameras or recording techniques for this?
The standard for optical sensors today is generally very high. The challenge with our applications is rather: How quickly do I need the image? And how quickly do I need the result in order to control the process? This means that speed is the main driving factor for us.
In addition to optics, artificial intelligence (AI) plays a major role here. How is this actually used in the systems?
In short: AI makes things easier for the user. This means that the user does not need to know anything about image processing - all they have to do is provide the image material. Here's an example: I have 500 images of cardboard boxes in good quality, send them to the AI and tell it: "If it looks like this, it's good." The AI develops the algorithm for image recognition itself. This is a game changer: much larger user groups can now use this technology - the entry threshold has become much lower.
Is the AI developed at ivii or are existing models used?
Basically, we use existing neural networks. What we do here in-house can be summarized with the term "hypertuning": This means that we tell the AI exactly what it should look at and what it can ignore.
Are your systems mainly used in intralogistics or are there other sectors for which they are of interest?
In addition to intralogistics - which is where we come from - our focus is on the manufacturing industry. In other words, where people assemble things. We can guide employees and confirm every step of their work.
Do you only target large companies with your products or are they also of interest to small and medium-sized enterprises (SMEs)?
SMEs are a very interesting target group for us, as they generally do not yet have quality assurance systems in this form. These companies also do not want to become dependent on a particular software house when they introduce such a system. With our solutions, they remain completely independent - they can learn everything themselves and adapt it to their needs.
The smartdesk supports employees in a variety of ways: by providing guidance, recognizing parts and confirming each individual work step. There are use cases from various industries. Can you tell us about them?
What the different sectors have in common is that they all had a quality problem that needed to be solved. This is because all companies have to deliver the highest quality in the shortest possible time, but do not have enough skilled personnel available to do so. This was the case for the logistics company Jerich, for RIKA Blechkomponenten GmbH and for Pankl's gearbox assembly. Our intelligent smartdesk assembly table provided the solution: at Jerich, left and right-hand holes in components are recognized for just-in-time delivery to the automotive industry and thus the component is correctly assigned. A person who does this all day long is bound to make a mistake at some point - not with the smartdesk. Or: when punching holes in sheet metal parts, errors can occur because, for example, the punching tool is broken. Here, too, it is impossible for a human being to recognize whether all the holes required for assembly are really present. The AI integrated into the smartdesk recognizes this. The intelligent workstation has also proven its worth in gearbox production: At Pankl, the "cognitive assistance" offered to humans by the smartdesk has even achieved a zero-error rate - and that with largely inexperienced employees, as there was a lack of specialist staff to increase shifts at the time.
What exactly do you mean by "cognitive assistance"?
We are all familiar with the physical assistance provided by technology: lifting, drilling, screwing, etc. We are now supplementing this physical assistance with "cognitive assistance": in other words, we support people with mental challenges as part of their work. And this support is gladly accepted, as employees can be sure that they have done everything right when they go home from work in the evening.
A special feature of the smartdesk digital assembly table developed by ivii is the simple sharing of knowledge. How does it work?
When developing smartdesk, the focus was on enabling customers to prepare the system themselves during work preparation. This means that the customer does not need a software company to train the smartdesk during a production changeover - they can easily do this themselves using their existing skills. This means that the company remains independent and flexible and can react quickly to changing market requirements. What's more, what has been trained once can be used immediately at other workstations - even at other locations. In this way, the smartdesk makes a significant contribution to knowledge scaling. Existing assembly workstations can of course be retrofitted with our technology.
Finally, where does the unusual company name ivii actually come from?
Originally, this was an acronym for Intelligent Visual Image Identification. We no longer use this today - but this is how our company name came about in 2016.
