If there is one sector that is changing rapidly, it is data and AI. New tools are emerging at a rapid pace and customer demands are changing accordingly. What is innovative today will feel outdated tomorrow. That is precisely why Pipple actively collaborates with European universities and research institutions. The MAGICIAN innovation project clearly demonstrates why this is so valuable. In this project, we are working with eleven partners from seven countries to develop technology that automatically detects and repairs defects in production processes.
Leading in innovation
Geert Driessen works as Data & AI Consultant at Pipple. Since MAGICIAN started over two years ago, he has been involved in this innovative project. ‘To really stay ahead as an organisation, you need to be close to research,’ says Geert. ‘MAGICIAN offers us that opportunity. By collaborating with researchers and experimenting with the latest models and methods, our team knows what is possible and we broaden our perspective. That gives us a head start, both for Pipple and our customers.’ In addition to Geert, three other Pipple colleagues are working on MAGICIAN. ‘We are broadly involved and constantly coordinate what efforts are needed.’
Intelligent robotic solution
MAGICIAN is a four-year EU Horizon project in which eleven partners are working together to develop an intelligent robotic solution. The initial focus is on superficial defects in the automotive industry. Using cameras, sensors and AI models, robots learn to recognise deviations and address them immediately. Later, the solution is expected to be more applicable across a wide range of production processes and industries.
Pipple joined the project because they were looking for a partner with strong AI and data expertise. ‘We are working on several components: from the configuration optimisation tool that makes implementation in other use cases easier, to detection models, route optimisation of robot arms, optimisation of the spot welding process and research into GenAI applications.’
A closer look: the detection models
Geert explains one of the components in which Pipple works: the detection models. ‘We develop Machine Learning models that recognise defects based on tactile data. We work closely with the partner who develops the tactile sensors. The tactile data comes from an accelerometer and a force sensor, which together transmit six measurements at a specific frequency. There is a lot of noise in that data, which makes it difficult to recognise small deviations. Nevertheless, we have managed to achieve a high degree of accuracy in a controlled environment.’
The next step is to combine these insights with image recognition. ‘By combining the results from both models, we can make an accurate prediction. Based on this, a Pipple algorithm determines which defects are given priority and which route the robot should follow.’
Where theory meets practice
Pipple is one of the few parties in MAGICIAN that is not a university or research institute. That gives it a clear role. Geert: ‘We not only contribute AI knowledge, but also ensure that solutions work in practice. We translate theory into tools that are compatible with the real production environment. This prevents us from getting stuck in what works well “on paper”.’
International collaboration
The project brings together partners from different countries, such as Italy, Sweden, Greece, and Turkey. This generates energy, as well as unique insights. Geert: ‘The cultural differences are sometimes greater than expected. In the Netherlands, usually, an agreement is an agreement. Elsewhere, it tends to be more flexible. And while we typically leave work at 6 p.m., in Southern Europe, a meeting might just be starting. It is precisely these differences that make the project enjoyable. You learn how other teams tackle problems, and that broadens your perspective.’
The impact for Pipple: knowledge, growth and a competitive edge
MAGICIAN brings Pipple a lot of value:
- We remain at the forefront of AI innovation. By working closely with researchers, we discover new techniques before they become widely available.
- We deepen our expertise in the manufacturing industry. This helps us to tailor our solutions as closely as possible to practical applications.
- We learn and develop internally. By participating in an international setting and testing new methods, our colleagues gain valuable experience that we can later apply in customer projects.
Integration and testing
The project is currently halfway through. The coming period will focus on testing, improving and further integration. Geert: ‘The initial results are promising: we have a better understanding of what works and what doesn’t, and we can see how the individual components are increasingly coming together to form a single intelligent robotic solution. The first major tests are now starting in Turkey and Italy, integrating a complete robotic solution with two robotic arms in a production environment. That’s where we’re also testing our software modules.’
For Pipple, MAGICIAN feels like a project that perfectly matches who we want to be: innovative and deeply rooted in data and AI. At the same time, it keeps us focused on the future.
Want to know more about the project?
Visit the official website: https://www.magician-project.eu/
