The project was one of 16 others that applied for implementation by the leading-edge cluster it's OWL in a call for proposals by the state of North Rhine-Westphalia; seven projects were selected by an independent panel of experts. The AI project in the area of intralogistics prevailed in the selection process for good reason: The potential that lies in the optimisation of timing, coordination, and organisation of delivery processes in and around the company is great: processes in warehousing, transport and order processing are to be optimised by at least 10 to 15 percent. Minister of Economics and Innovation Prof. Dr. Andreas Pinkwart: "In the selected projects, industry and science are jointly developing new technologies for the digital transformation in SMEs. The partners are addressing the needs of industry and developing practice-oriented approaches and solutions that other companies in OstWestfalenLippe and throughout North Rhine-Westphalia can use. In this way, they make an important contribution to a competitive industry in North Rhine-Westphalia."
 

What does AI have to do with logistics? 

This is the initial question to ask when looking for the potential of machine learning or artificial intelligence for logistics processes in companies. Intelligent algorithms and machine learning methods are first of all able to process the large amounts of data that are present in logistics processes. In particular, we are talking about inventory data, production input and output, ERP data as well as resource availability and occupancy plans of machines and systems. In the first step, processes and workflows are modelled and mapped graphically so that people can grasp and understand this aggregated data at a glance, for example to identify bottlenecks, delays or idle times. In the next step, the models can be built using machine learning techniques that can detect or even predict irregularities. 

The initial meeting of the project partners

took place on Thursday 18.06.2020 - in accordance with social distancing, project promoters and partners met in digital form, as did the new head of department at Fraunhofer Lemgo in the field of machine intelligence. Dr Oliver Niehörster and the AI expert Kaja Balzereit, M.Sc. took part. The industry partners used the round to present their own use cases and ideas and to point out the different motivations, goals and needs for participating in the project. Here, the concrete challenges of their own operational logistics became clear. The use cases vary depending on the use of technologies and the industry, just like the goals of the companies in the project: "Some project partners focus on monitoring intralogistics, others are interested in optimising specific processes or routes, others want to be able to dynamically adapt the distribution of goods and resources and make it more resilient to disruptions. Before these goals can be operationalised, the sensor and process data must be shaped into a model using artificial intelligence," explains Kaja Balzereit.

The roadmap of the Fraunhofer scientists roughly envisages four steps for the development of the AI instrumentation: 

1. Scientific requirements analyses: Here the researchers examine the state of the art and elicit the available machine learning methods. In particular, the question of the extent to which these need to be configured and rewritten for the respective use cases is interesting here: 
2. Data and IT architecture: An important part of the project - as in almost all AI-related research projects - is the preparation and homogenisation of the data sets. This "homework" is necessary to ensure availability and compatibility in the first place. 
3. KI library: In the course of the project, the tested and processed procedures are to be compiled in a collection. This "AI library" is to bundle the tested, tried and tested procedures and make them available even after the project has been completed, so that they could also be further developed and applied in other companies and across sectors in the future.
4. Empowerment: This project phase provides for sensitising and educating employees of the companies for the use of AI. This is where Fraunhofer's years of research experience in the fields of machine-human interaction and the implementation of socio-technical systems in the industrial environment come in handy. "In all innovations in the production environment, the social, corporate cultural and psychological issues that matter are those that place the workforce at the centre of the considerations. If the people in the factories understand what the processes are based on and how they are used, reservations are quickly dispelled. For us, it will then be an exciting task to deepen or transfer the experience with AI from other production areas in the intralogistics sector, for example to the freight forwarding sector. In our estimation, AI offers the entire spectrum of logistics numerous opportunities to make processes simpler and more efficient, as well as to open up new business models," Balzereit continues.