Automated Driving Partners

ANITA: triple interpreter waiting in the starting blocks

In the ANITA project (Autonomous Innovation in Terminal Operations), MAN Truck & Bus wants to make a fully automated truck a reality in terminal operations between the DB Intermodal Services container depot and the DUSS terminal in Ulm Dornstadt. Hochschule Fresenius University of Applied Sciences, as project partner, has already completed its scientific system analyses for this project in the summer of 2021.

Now the researchers are announcing the next milestone:

The software for mission planning is ready, so it can now be linked to vehicle development. In an interview, Professor Dr. Christian T. Haas from Fresenius University of Applied Sciences and Anselm Pilz from Deon Digital explain the details: What exactly did the researchers and developers program?

In the summer of 2021, you completed your scientific analyses for ANITA. What data did you collect at that time?

PROF. DR. CHRISTIAN T. HAAS: If a truck is to move automatically from one terminal to another, as planned in Ulm, delivering and picking up containers, a lot of communication happens and with each communication a decision that the truck will have to make in the future without human support. We have to ensure that the communication chain is fully mapped and that the right decision is always made. That's why we analyzed every conceivable situation to capture the digital, unambiguous information it needs. After all, the truck only has a limited pool of information, unlike humans, who can get further information themselves if in doubt.

What has happened to these results since then?

PROF. DR. CHRISTIAN T. HAAS: From this work, i.e. from the realization of where the truck has to make a decision depending on the IT systems of DUSS and DB IS, we have created a module plan. In it, three different worlds exist: the IT system of the truck, the IT system of DUSS, and the IT system of DB Intermodal Services - the last two having been developed independently of each other over decades and "speaking completely different languages." However, the truck needs information from both systems for its automated work. To ensure that all three systems can communicate with each other without errors, we have built modules that link everything together - in the form of various protocols and interfaces. In other words: We programmed an interpreter that can read each system and translate each unambiguously for everyone involved, so that no misunderstandings arise. It's as if three people at a conference want to talk to each other who can only speak French, Arabic and Chinese respectively and can now communicate with each other thanks to a triple interpreter.

So a kind of universal language ...

ANSELM PILZ: Exactly. If there is a translation problem, the truck stops or makes a mistake. For example, if the truck receives the signal to load a container, but it hasn't even unloaded its current container yet, it won't work. To avoid such situations, we used CSL code as our language, our Contract Specification Language. With this, all processes are set up as individual "contracts": The container may only be loaded onto the truck when the previous one has been unloaded. But then it must also be loaded.

The programming was done by Deon Digital as a partner of Hochschule Fresenius University of Applied Sciences. How did the collaboration between the scientists and the developers go?

PROF. DR. CHRISTIAN T. HAAS: Excellent. Even during our on-site analyses, there was an intensive exchange with Deon Digital throughout. During the programming work, we were in contact at least every two days, and we also held an intensive workshop together at Deon Digital's research department in Copenhagen that lasted several days. During this workshop, we visually prepared and presented all relevant phenomena in analog and digital form, subsequently translated them into programming code, and then translated them back for control purposes, i.e. the individual commands were reproduced and acted out with miniature MAN trucks.

So the programming of the circuit diagram is finally completed?

ANSELM PILZ: The programming and thus the "mission planning" stands - for the "happy flow", i.e. for the routine standard sequence as it is ideally planned. However, since there are also external influences and thus always unpredictable events, we also have to take this into account. We are currently developing solutions for this "unhappy flow". With the current status of the software, however, MAN and Götting can now already drive vehicle development forward.

Is software ever a "finished product"?

PROF. DR. CHRISTIAN T. HAAS: It will always remain an open solution. Because systems like the one in Ulm are networked worldwide and change. Suddenly, a new player can come along - be it international, regional participants or, for example, new technologies or means of transportation. So we have to take that into account in the programming so that the system is future-proof for new players that we have to be able to dock.

Is ANITA a special use case for Deon Digital as a practical use case from logistics?

ANSELM PILZ: Our language CSL was originally built for financial and insurance transactions, which mostly deal with virtual "goods" - even though CSL has already found use in other contexts. At ANITA, high technology in the form of CSL now meets very physical processes and traditions of the transportation industry. Massive containers as well as autonomously driving trucks cannot be moved virtually like money flows. That is already something special for us.

The Fresenius University of Applied Sciences has also conducted analyses at other DB transshipment points. Can the mission planning that has now been programmed also be used there or only in Ulm?

PROF. DR. CHRISTIAN T. HAAS: We assume that we will be able to transfer a large part of the programmed modules to other sites with the identical infrastructure as in Ulm. We must be able to understand the differences at relatively short notice and be able to adapt them through new programming code. The knowledge required for this is provided by the terminal staff, who must be enabled to quickly and clearly make transparent the respective phenomena that differ.

Let's take this one step further. Will all these analyses also be necessary for all other Hub2Hub applications in logistics in the future?

PROF. DR. CHRISTIAN T. HAAS: As soon as the automated truck is on the highway, there is a clear set of rules in the form of the Highway Code as to what a vehicle may or may not do there. These are relatively thankful boundary conditions for system control. It is comparatively difficult at transshipment points, where there are fewer clear orders and where rules have become established over time that are not explicitly recorded anywhere. We believe that "computational thinking" can be a game changer here: We need a simple language via abstraction of the processes, which at the same time ensures that the content is not distorted. But yes, at transshipment points this "translation" will always be necessary so that the individual systems can talk to each other.

What do you expect when the first test drives take place in Ulm as part of ANITA at the end of 2022, beginning of 2023?

PROF. DR. CHRISTIAN T. HAAS: We are already simulating jobs with the help of "digital trucks" to check the communication between the individual systems. This is a great help for the continuous optimization of programming, to be able to look into the future in this way, but also into the past. During the testing in the infrastructure in Ulm, we as Fresenius University of Applied Sciences and Deon Digital will act in two teams: One team will log the technical processes, while the other team will observe the process quite abstractly from different positions to see if there are any phenomena that we have not yet taken into account. Only afterwards will we then combine the results of the two teams.

What appeals to you personally about ANITA?

PROF. DR. CHRISTIAN T. HAAS: The first thing I find exciting is the large gap between the physical, repetitive and also system-relevant processes on the terminal site on the one hand and the so different culture of high technology on the other: In the transportation and logistics world, many processes are self-organizing or based on implicit - that is, not formalized or data-based - decisions. These are uncomfortable and sometimes incomprehensible situations for programmers; the programming world is based on a complete and precise data image of the real world. Developing solutions for all sides precisely in this area of tension is very exciting, but also not always easy. What also fascinates me about ANITA is that we are implementing the new technology while the system continues to run. In the long term, another interesting question is how we can map mission planning processes smartly. Here, we are currently looking at a mix of deterministic control of core areas and outsourcing processes that are capable of self-organization. Similar forms of organization exist in the human nervous system: The brain outsources simpler processes - such as reflexes - to the level of the spinal cord so that it is less stressed. It is not just a matter of the system functioning in principle, but also of how well it performs and how efficiently. However, there are still a few nuts we have to crack.


Prof. Dr. Christian T. Haas and Anselm Pilz in conversation with employees from Hochschule Fresenius and Deon Digital at a workshop

heads the Institute for Complex Systems Research at Hochschule Fresenius University of Applied Sciences. He works with his scientific team at the House of Logistics & Mobility in Frankfurt am Main. Together with MAN and Deutsche Bahn, he has already implemented the successful platooning field test EDDI. In the follow-up project ANITA (Autonomous Innovation in Scheduling), the systems researchers around Christian Haas are now working on implementing automated hub-to-hub transports.

works as a Business Engineer at Deon Digital, a software development company founded in 2017 with headquarters in Zurich, development office in Copenhagen and additional offices in Germany and Hong Kong. Deon Digital has focused on projects from scientific research and supports Fresenius University of Applied Sciences at ANITA. Deon Digital's CSL Contract Specification Language is used as the programming language.