All Stories
9 min

“KION has Already Seen Tangible Successes for Real-World AI Applications”

Whether it’s smart factories, assistance systems, or automated guided vehicles (AGVs) – the KION Group’s brands are increasingly deploying artificial Intelligence (AI) systems in their solutions. AI will help optimize warehouse operations, help anticipate demand and bottlenecks, and reduce the risk of overstocks or shortfalls. Harnessing our position as a global technology leader, the KION Group is proactively cultivating the potential that AI offers.


Joachim Tödter, Senior Director Technology & Innovation, has been instrumental in shaping and driving the company forward in this area for a number of years. He knows: The entire industry is facing a paradigm shift away from rigid systems towards dynamic, open, and AI-driven processes. Neural networks are also growing in importance. He explains what this means in this interview.

Mr. Tödter, when did you start exploring the future of KION?

Around 20 years ago, I worked in advance development. Back then, there were a number of forward-thinking sales colleagues that were working in close cooperation with customers and they firmly believed that the future lies in the realm of flexible automation. I took their ideas seriously and, very early on, we started creating future-oriented solutions. As early as 2008, we supplied a customer with a pilot project that was well-ahead of its time; an adjustable rack system with shelf heights of up to eight meters that utilized hybrid operation, combining manual and automated processes.

So, has automation advanced since those early days?

Yes, it’s important to understand that Automated guided vehicles have existed in some form for around 30 years. However, commissioning each system represented its own unique challenge and any modifications to these systems required considerable configuration. This is where we come in: Advanced sensors and sophisticated hard- and software solutions make vehicles more flexible in their deployment. In the future, we’ll be able to completely do away with the initial setup process because forklift trucks will be able to adjust and orient themselves autonomously. As warehouse inventories change, the machine is able to recognize this and “think” accordingly. The trend is moving away from fixed automation to increased flexibility. A good example of this is STILL’s involvement in the “IMOCO under Industry4.E” project. This project is aimed at fitting AGVs (automated guided vehicles) with state-of-the-art sensors and AI to enable them to move autonomously through the warehouse and avoid any obstacles. Cameras, radar, and laser scanners play a crucial role in this process, and, with the help of artificial intelligence, AGVs can classify surrounding objects.

The project IMOCO , which is supported by the German Federal Ministry of Education and Research and the EU, is aimed at developing intelligent autonomous forklift trucks for use in production environments and warehouses. It includes the use of artificial intelligence and modern sensor technology and communication to allow forklift trucks to navigate independently, to avoid obstacles, and to intelligently find their way through the warehouse. The project is scheduled to run until Q4 2024.

IMOCO: Making Autonomous Transportation Vehicles More Intelligent.

Machine vision is the key word here, right?

Yes, definitely. Cameras, combined with AI-based image analysis, means AGVs are advancing to the point where they can autonomously “identify” objects, i.e. what’s a shelf, what’s a person, and what’s a pallet. They can differentiate and recognize environmental conditions and interpret and react appropriately – that’s the key. Dematic’s drones for analyzing storage and retrieval systems are just the first example: Machine intelligence will soon be able to recognize errors or damage to these systems. That’s a completely different ball game.

How does this work in practice?

The core principle revolves around neural networks – the system’s ability to autonomously learn new skills. Neural networks have to first be trained on data input before they can put this learned knowledge into practice. In terms of image analysis, this implies the ability to recognize objects that resemble the training examples. It’s a little bit like teaching a child: You show a child pictures of cats and dogs until the child is able to differentiate what’s a cat and what’s a dog for themselves. The difference is that AI is able to learn much faster than a child.

A neural network is a machine learning model inspired by the structure and function of the human brain. It is capable of learning from data sources and is systematically trained using that data to improve its performance. A complex network of connections arises from the interaction of individual neurons, which is able to perform predictions and classifications, as well as recognize patterns.

What are most significant opportunities for the use of AI?

This technology represents myriad opportunities, from enhancing efficiency through to increasing flexibility – so much is possible. For example, AI systems can help optimize inventory management and help avoid overstocks or shortfalls. They can utilize preventive data analysis to identify potential system issues ahead of time or anticipate energy spikes to help save money, as has already been proven at Linde Material Handling through its collaboration with ifesca. Another example is the “ARIBIC” research project started by the KION Group and its partners. This acronym stands for Artificial Intelligence-Based Indoor Cartography and the project has helped demonstrate what is possible when we create a digital twin of an entire warehouse using data, providing a real-time visualization of operations and processes in the warehouse.

The ARIBIC research project is aimed at creating a real-time digital twin of a warehouse. KION is working with LeddarTech, the Karlsruhe Institute of Technology (KIT), and the STARS Lab at the University of Toronto for the project. ARIBIC is able to create high-resolution 3D maps of the warehouse by systematically collecting and processing data from sensors equipped on autonomous guided vehicles. These maps display real-time information that can be used for simulation processes and to optimize driving routes.

Do you think that AI will develop to the point where it’s as intelligent as a human within the next 10 or 20 years?

We first need to clarify what ‘intelligent’ means when it comes to humans. When we consider what AI solutions like ChatGPT can already achieve today, we can see that these solutions have developed much quicker than even experts imagined. I can remember a pilot project where we tested machine vision on forklift trucks. A few years ago we saw recognition rates of 80–85%, which was quite remarkable but impractical for real-world applications due to the high error rate. But things have moved on to the point where recognition rates are so good that you don’t even have to think about whether AI implementation makes sense – it’s a given. AI’s capabilities are constantly improving, and fast. We can now see that the capability of AI’s is doubling every eight or nine months, and this is coming down all the time. We’re already using AI in R&D to help write software!

Can you see an AGV being able to autonomously develop solutions for intralogistics challenges at some point in the future?

Absolutely! Let’s look at the example of reinforcement learning, where a virtual model is presented with a challenge that it needs to find a solution for. We’re already doing this when it comes to analyzing how an AGV should pick up a load, for example. We provide the machine with a simulation environment where it can try out countless examples in any shape or form until it’s able to determine the most efficient solution. The happens in a completely virtual space, overnight, and thousands and millions of times until it finds a solution that meets all challenges. Engineers no longer have to program the machine with countless routes, it’s able to autonomously work out the solution.

Machine vision is a branch of AI that allows computers to interpret visual information from images or videos. It’s based on neural networks that process image data, analyze the images, and recognize patterns and characteristics. These can then be used to identify objects, people, scenes, and much more.

Will we still need people in the warehouse of the future?

From what we can see, people will still be needed in the warehouse, but there are countless monotonous and physically demanding tasks in warehouses and our customers are already struggling to find employees to complete these. This is why we’re seeing an increase in demand for automation. AGVs will replace drivers for these tasks. But we won’t be able to replace drivers as quickly for all tasks that demand a delicate touch or require highly adaptable execution. Machines will replace the human factor in many industries for repetitive tasks, but specialized employees who can install systems, test them, and keep them running will be more in demand than ever before.

Is there a KION product which you can say represents a shining example of the intralogistics of tomorrow?

Automated guided vehicles such as STILL’s iGo neo will become an integral part of intralogistics in the near future; an intelligent product that works with people to simplify processes for the operator. The ifesca load management project is very promising, as is the LoadRunner which utilizes swarm-enabled logistics for transporting multiple small containers. As is the ARIBIC research project, which makes it possible to see real-time inventories. So you see, we’re not in short supply of shining examples at KION.

By means of the latest sensor technology, the OPX iGo neo detects its operator, its surroundings, obstacles and distances.

Can you see AI making your job obsolete at some point in the future?

Absolutely – maybe in 30 years, maybe in 300. But that doesn’t mean that the people of the future will be able to just spend their time lazing in the sun. We are currently addressing the concept of “weak” AI, which will undoubtedly function remarkably well and help make our lives easier. I currently find it hard to believe that a “strong” AI will be developed at some point in the future that will be able to replace humans across the board. I see enormous potential for “weak” AI, which works with people and handles repetitive tasks. It’s coming, and it’s going to help us. I can’t currently see humans losing control of these systems, even though it’s interesting, and absolutely vital, to think about the risks and opportunities of “strong” AI.

The keyword here is risks, what are those?

The fear of job losses is a legitimate concern. As well as autonomous mobility, there are certain aspects of warehouse logistics that AI-based automation solutions will increasingly be able to meet. Automated warehouse management, which KION subsidiary Dematic already offers for example, will play an increasingly important role in the future. This allows us to automate processes such as inventory management, shipping, and order picking. But this also creates new jobs relating to the maintenance and operation of AI systems but some professional fields will change dramatically in the coming years. But this doesn’t just apply for intralogistics. I’d like to share the findings of a recent Goldman Sachs study from 2023, which suggests that up to 300 million jobs will be affected by the latest developments, in particular for those in management, in software programming, or the creative sector. It’s definitely food for thought.

LoadRunners pick up parcels completely autonomously and put them back down in the right place without colliding with each other.

And what about data security? AI systems require large amounts of data to be able to work.

That’s right. And that’s why we’re developing KAP, KION’s Analytics Platform, where we can store this data securely but where authorized personnel can easily access this information. Data protection is essential, and it encompasses not only the quantity of data but also the need for robust security measures. Data quality is also key. It can cause significant problems if AI systems make the wrong decisions based on imprecise or incomplete data. This is where humans remain key, AI systems need to be regularly monitored and validated to ensure that they work correctly and make the correct decisions. And in this context, humans will remain indispensable in the long run.