Fraunhofer Institute for Material Flow and Logistics IML
Fraunhofer Institute for Material Flow and Logistics IML

Artificial intelligence causes a paradigm shift in logistics

The development of artificial intelligence (AI) is unstoppable. Intelligent systems are increasingly controlling processes and making decisions that influence everyday life. Although the future technology has already entered our daily lives, we still face great challenges. In order to meet these challenges, Fraunhofer IML has entered the ring with the “Social Networked Industry,” researching and developing AI systems for human-machine interaction. Because one thing is certain: AI is determining the future of logistics.

The most recent examples, such as ChatGPT from the Microsoft subsidiary Open AI, are impressive. They show what machine learning methods and algorithms can now achieve and how incredibly fast they are developing. This is a result of generative artificial intelligences of the third generation. Unlike the rule-based systems of the first generation and the learning-capable programs of the second generation, these are able to autonomously combine knowledge and sometimes context as well with data. 

AI as a key technology

The third generation of AI offers great opportunities, especially for complex tasks in logistics. For this reason, Fraunhofer IML has been using artificial intelligence (AI) and open source to research innovations regarding the current megatrends for some time now. As part of the Lamarr Institute for Machine Learning and Artificial Intelligence, the Dortmund institute wants to further develop the third generation of AI, or “triangular AI³,” as the researchers call it, in a multidisciplinary way. This is the ultimate technology for developing future technologies for the logistics sector.

The researchers believe that deep learning will make third-generation artificial intelligence become the actively operating part of logistics and production systems in the future, since it can solve more and more complex problems in addition to routine tasks. Deep learning programs are now probably even superior to human intuition – especially in the case of larger data volumes. As a result, these programs would also potentially optimize all processes along the supply chain. In this way, AI could become an actively engaged partner of human beings. It would not only provide information but also be able to control, negotiate and plan more and more actively. It could be used to solve the great problems of our time – starting with geopolitical crises and disruptions in supply chains, to labor shortages and demographic change. “We can already manage some challenges in logistics better with artificial intelligence, and some problems we will even only solve using AI. Whether organizing a swarm of mobile robots, combing through large databases or calculating the next batch, many things in intralogistics cannot be sufficiently described with formulas or controlled with common sense. Here, AI can help and learn some things that we do not understand,” explains Prof. Michael ten Hompel, executive director of Fraunhofer IML.

Value-based application of AI

To increase the acceptance of AI-based solutions, an important focus of their research is making machine learning methods transparent and interpretable for people. Decisions and mistakes of AI must be seamlessly traceable. Trustworthiness also means complying with ethical and data protection standards, however. With the slogan “Trustworthy AI,” the scientists have therefore focused on an important principle for developing AI: How can machines be programmed so that they act “responsibly,” in other words, according to standards and rules? After all, AI should serve humans and not vice versa. As a basic technology, it is supposed to be a helping hand to support people cognitively as well as physically. The institute wants to use AI to holistically establish safe and healthy work practices in process automation and process autonomization, from the shop floor to the value-added level. In addition, existing highly scalable digital and technical solutions for applications are to be systematically interlinked and developed further for all company sizes. Fraunhofer IML wants to create an open-source ecosystem for logistics – a type of “Linux for logistics and AI” – with the companyfunded Open Logistics Foundation initiated within “Silicon Economy.” The institute wants to use this to pave the way for AI-based technologies for small and medium-sized enterprises as well. Together with the “Social Networked Industry,” they want to create a world in which humans and AI cooperate as partners in social networks and thereby shape the world together. 

“Social Networked Industry” – a foundation for secure and trustworthy interaction between humans and AI

KI und Mensch Interaktion
© Fraunhofer IML

With the project “Social Networked Industry – secure and trustworthy cooperation between humans and artificial intelligence,” Fraunhofer IML wants to use AI to create a new type of working world for logistics with humans at its center. Humans become the directors of entire systems and interact with intelligent and networked machines. This results in social networks that connect humans and technology.

Fraunhofer IML launched the “Social Networked Industry” in 2017. The researchers already anticipated that AI would massively change the world. “Digitalization of everything and artificial intelligence in everything will change everything for us. A social networked industry will arise in which humans and machines work together as partners in social networks based on artificial intelligence,” Prof. Michael ten Hompel envisions. 

How is the Social Networked Industry defined? 

The “Social Networked Industry” consists of two components: the networked industry and the social network. Topics such as Industry 4.0 and the Internet of Things deal with the networking of technologies. However, networking is also a significant factor in today’s society. Social networks such as Facebook, Twitter, Xing and the like connect people with each other across national and cultural borders. They are decentralized, intuitively usable, scalable, emotional and have a permanent place in our everyday lives. Networking also plays an increasing role in companies. For this reason, the researchers want to transfer this concept to the cooperation between humans and machines. The “Social Networked Industry” stands for a combination of autonomously acting human and machine entities that grow into functioning dynamic networks. These can arise on an ad-hoc basis to solve a concrete problem. The spontaneous networking and fast creation of new structures are essential both within companies as well as across company boundaries. This is the only way that the logistics world can master future challenges and optimally exploit opportunities to leverage competitive advantages. According to the researchers, human-machine communication allows the positive unique characteristics of humans and machines to be combined: flexibility and creativity with efficiency. The aim is to relieve people from error-prone and monotonous activities and increase their satisfaction. This requires people to be trained, because the technical and organizational transformation will change activities and job profiles. Lifelong learning is therefore essential to prepare people for their new roles. 

Demonstration platform for application-related developments

A research and demonstration platform for applicationrelated developments for logistics is also to be set up within the “Social Networked Industry.” Real scenarios and conditions in operative logistics are to be simulated. This will test and evaluate innovative technologies as well as the cooperation between humans and technology. For example, the Embodied AI department of Fraunhofer IML, which deals with robotics, has already been using a new generation of simulation-based AI for quite some time. This is used to develop autonomous solutions such as the highly dynamic robot “LoadRunner” or the transport robot “evoBOT.” 

The simulation is so exact that it can be used to teach very complex situations. Sensors in the simulation are taught to locate vehicles in a swarm. The connection with control technology produces a digital continuum – a self-optimizing system.

Flughafen München Zukunft evoBOT
© Fraunhofer IML

The “Social Networked Industry” is a task for society as a whole

The vision of a human-centered “Social Networked Industry” has to be understood as a task for society as a whole, in which experts from various specialist disciplines work together on equal terms to design it. The graphic on p. 8 shows which research questions have to be addressed for this purpose

Robot colleagues – How humans and technology become a team

Flughafen München Zukunft evoBot
© Fraunhofer IML

For Fraunhofer IML the cooperation of humans and technology is one of the great challenges of the digitalization process. It affects both the quality of work as well as people’s acceptance of digital solutions. Digital assistance systems that support people individually and ergonomically with their work could be the solution here.

When selecting and designing assistance systems, it is therefore important to find the right design for the respective process and the involved employees. The researchers want to find out what information the employees need in the process and how this information can be provided. For this purpose, they study different types of assistance systems, such as AR and VR goggles or exoskeletons, for example, taking into account the basic rules of cognitive ergonomics. According to Fraunhofer IML, assistance systems are also able to support people as data-based decision-making aids in planning and implementing logistical processes. Simulations and artificial intelligence based on previously recorded data can provide a valuable impetus here. 

Fraunhofer IML wants to reduce health risks such as stress and posture problems, increase the acceptance of technological innovations, as well as boost the efficiency of work processes through so-called “social machines” using human-technology interaction. This will make people’s interaction with autonomous drones and driverless transport systems as well as the work with virtual reality (VR) and augmented reality (AR) increasingly significant as well.

Privacy warning

With the click on the play button an external video from www.youtube.com is loaded and started. Your data is possible transferred and stored to third party. Do not start the video if you disagree. Find more about the youtube privacy statement under the following link: https://policies.google.com/privacy

Machine learning and neural networks fuel development of artificial intelligence

Artificial intelligence imitates human cognitive abilities by recognizing and sorting input data. This can be based on programmed sequences or generated by machine learning. With machine learning methods, an algorithm autonomously learns to perform a task through continuous repetition. The machine focuses on a specified quality criterion and the informational content of the data. Unlike with conventional algorithms, no solution path is modeled. The computer autonomously learns to recognize the structure of data. For example, robots can learn by themselves how to grip certain objects in order to transport them from A to B. They are only told from where and to where they should transport the objects.

Flughafen München Zukunft evoBOT
© Fraunhofer IML
KI Box in Lagerhalle
© Fraunhofer IML

The robot learns how to grip by repeatedly trying to do it and receiving feedback from successful attempts. The continuously growing amount of data and the technical progress of computing power have helped to make increasingly complex calculations possible using machine learning. As a subfield of machine learning, neural networks consist of artificial neurons. These use algorithms to imitate the nerve cells of the brain. Like nerve cells in the brain, the artificial neurons are interlinked and process information through deep learning. As a result of training with large data quantities, neural networks can learn to recognize patterns and relationships as well as to make predictions. They are able to improve themselves. Neural networks and deep learning are used, for example, in image and speech recognition, automatic translation, prediction of behavioral patterns and automatic decision-making.