Logistics, like geometry, is completely deterministic and algorithmizable, states Prof. Michael ten Hompel, Executive Director of the Fraunhofer Institute for Material Flow and Logistics IML. This is because all process steps in physical logistics, including picking, transport, storage and retrieval, and handling, are already fully described. The combination of these process steps results in highly complex optimization problems. This also makes logistics particularly well suited for the use of (self-learning) AI algorithms.
External Expertise is Crucial
A 2020 study by Tata Consultancy and industry association Bitkom shows that while nearly half of all companies (47 percent) see AI as a key technology for logistics. But only 27 percent are already using AI or have specific plans to use it. The main hurdles cited were high investment costs, challenges with data security and data protection, but also a lack of expertise.
So far, many companies are still hesitant to delve deeper into AI technology and machine learning, and SMEs in particular are finding it difficult to do so. This is due not least to the drastic shortage of specialists in the field of data science. In most projects, therefore, external expertise from consultants is called in. The head start that AI users gain is not so easy for other market players to catch up with, because it often involves learning processes and building up a qualified data basis.
Where AI Makes the Difference
There are many possible applications. In practice, it is not uncommon for goods to sit around on the ramp for a day because they were picked too early. Intelligent software in conjunction with IoT scenarios such as tracking can make it possible to plan much better today. For example, if geofencing is used to signal a truck’s arrival in an hour, goods can be staged on time and loading teams can be managed accordingly. This data can be used to train an artificial intelligence, for example, to make predictions about punctuality on certain days or times, as well as about the reliability of the respective carrier. As a result, for example, transports may no longer be booked on Tuesdays or from transport company XY.
The topic of picking is not always trivial, as the example of home24 shows well. At the online furniture retailer, the packages and thus also the loading and unloading times differ extremely. Machine learning can be used to predict very precisely how much time will be needed for each delivery and thus optimize team planning. In general, improving forecasts is one of the key benefits of AI technology.
Taking the Shortest Path
Route optimization in the warehouse generally still has room for improvement. Machine learning can be used to automatically calculate how often goods are picked, check how long they take to be made available, and deduce which goods should be stored in a more accessible location. The basis for this is, of course, qualified data, which can best be collected via data acquisition on mobile devices. In the best case, every process step is documented almost in real time. Camera images of the storage rooms and areas can also serve as a basis for calculating the best routes.
There is also a trend toward autonomous transport systems, in which AI ensures on the one hand that the environment is “seen” appropriately by sensors so as not to cause accidents. On the other hand, AI algorithms calculate the most efficient route in each case here as well.
Counting and Reading
But intelligent solutions also replace a great deal of manual work when it comes to taking inventory and counting certain goods or materials: for example, a specially trained image recognition algorithm can immediately detect how many steel pipes are stored on a load carrier.
However, AI can also be used in the sense of significantly improved text recognition: For example, by automatically recognizing the license plates of trucks from different countries and checking whether the truck has docked at the correct ramp. Label recognition on containers can also be automated to check whether the container matches the advised number, for example.
Outdoor Areas in View
Wherever material is stored in the yard that cannot be fully monitored, a combination of drones and image recognition can provide security and greater transparency. That’s because with AI image recognition, even small deviations or irregularities become clear. In the case of coal or steel, for example, the algorithms use the drone images to automatically recognize how high the inventory is and what tonnage of a product is present: This makes it possible to detect thefts when the data is compared with the information in inventory management – an added value that also has a direct monetary impact.
Latest news and blog posts from the world of intelligent supply chain management.