(Foto: istockphoto)
Back to overview

Autonomous driving

Develop together, test digitally

How digital twins and interdisciplinary partnerships between industry and science are driving development.

Lutz Morich has created a digital twin of Ingolstadt together with his team. (Photo: Audi)
Lutz Morich has created a digital twin of Ingolstadt together with his team. (Photo: Audi)

It is a perfectly normal town in Upper Bavaria, except for the fact that there are two of them: the real Ingolstadt and its digital twin. As part of the project SAVeNoW, Audi has created a digital version of the town where the company has its headquarters. The VW subsidiary is managing the project together with a host of scientific partners, including TH Ingolstadt, KU Eichstätt, TU München, German Aerospace Centre DLR, University of Stuttgart and the Fraunhofer Institute for Transportation and Infrastructure Systems IVI. Also involved are the following companies: Continental, TWT, ASAP, 3D Mapping Solutions, seppmed und Elektronische Fahrwerk-Systeme (EFS). The project is financed by a grant of 7.5 million euros from the Federal Ministry for Transport and Digital Infrastructure.

‘We have created a virtual test field for automated driving,’ explains SAVeNoW project manager Lutz Morich, ‘to answer questions that cannot be solved analytically.’ Morich and his team have been working on the simulation for over three years now. The simulation encompasses all static objects, from inner city streets to the A9 motorway, and from industrial areas to all the city’s buildings.

Virtual test field for autonomous driving

This representation has been enriched with dynamic elements, all kinds of vehicles and people. Data from various sources has been integrated, such as from traffic monitoring systems, digital traffic lights and networked Audi test vehicles, a large number of which travel around the city’s roads.

‘Autonomous driving has a different level of complexity on the motorway compared to urban areas. Autonomous driving in the city is the supreme discipline,’ emphasises Morich. ‘A city offers a great many more options and the number of possible scenarios is far greater here.’ How does a control algorithm on an autonomous vehicle react when in traffic with cyclists and pedestrians as well as to the more complex rules of city traffic? This can be best developed and tested using the digital twin. ‘Some people used to say that we could test this only on the road, but there are far fewer of those voices now,’ explains the expert. And other companies are also testing their technologies in a virtual environment like this, some of whom Morich is in pre-competitive contact with.

Acceptance by society

To find out how the algorithm reacts, an autonomous vehicle can be “inserted” into the virtual depiction of the real Ingolstadt traffic. But this depiction can also be deliberately influenced. ‘In an agent-based simulation, for example, I can simulate specific driver types as road users, such as particularly aggressive or particularly cautious drivers and then look at what the algorithm does.’

Simulations in the digital twin are, according to Lutz Morich, particularly important for three reasons. Firstly, for the testing that accompanies development, whereby each modification to the algorithm can be analysed. ‘In this way, we as manufacturers can guarantee the safety of our development quality.’ Secondly, it is a tool that can be used for approval and certification. Now that autonomous driving has started, regulatory authorities must approve functions in vehicles to be used on the road that have never existed before. ‘And to do so, they need the appropriate tools and expertise.’

The third reason might not be obvious at first glance, but it is very important to Morich nevertheless: acceptance by society. ‘We can show people the digital twin with the autonomous vehicles, even let them experience autonomous driving and therefore assuage their fears,’ says Morich. ‘This is something that we have actually already done in Ingolstadt’s pedestrian zone and we got some very positive feedback.’

Timo Woopen leads the UNICARagil project, a collaboration between numerous partners from research and industry. (Photo: RWTH Aachen)
Timo Woopen leads the UNICARagil project, a collaboration between numerous partners from research and industry. (Photo: RWTH Aachen)

Four autonomous vehicles based on new architectures

The UNICARagil inititativealso uses simulations to assist development. ‘This reduces the expensive testing time spent on the vehicle,’ explains project leader Timo Woopen from RWTH Aachen.

The project was launched in 2018 with a grant of 26 million euros from the Federal Ministry of Education and Research. Upon completion of the project in 2022/23, the aim is to present four autonomous vehicles. These will comprise product prototypes from the shuttle vehicle, taxi, cargo and private car segments. Project manager Woopen is confident that the deadline can be met. ‘Three of the vehicles are already on wheels.’

A modular vehicle concept was developed during the project that comprises six architecture levels. These range from the physical vehicle geometry and software to the networking architecture. There is also a strong focus on electronics.

Project with over 100 scientists

UNICARagil is arguably a unique project in this field. Over 100 scientists from 16 chairs at 8 universities from Braunschweig to Ulm have joined forces. Various disciplines are represented, from automotive engineering and informatics to psychology and satellite geodesy (satellite-based surveying of Earth). An additional eight partners from industry are also involved, mainly SMEs and start-ups.

‘Working in such a broad consortium is great fun, but unusual,’ emphasises Timo Woopen. ‘Universities are normally part of the competition, for example for research funding.’ Why are there so many collaborations working on autonomous driving in general? ‘It has to do with the market,’ explains the leader of the Vehicle Intelligence and Automated Driving research unit at the Institute for Motor Vehicles at RWTH. ‘German companies are very good at building traditional cars. But when it comes to software, players such as Google or Amazon enter the market with a lot of money and expertise.’ So by combining forces, he says, we have better chances.

When autonomous driving will come

And when will autonomous driving be part of everyday life? Lutz Morich and Timo Woopen are reluctant to make any predictions. But they are agreed on the fact that the German law on autonomous driving, which was passed at the end of July 2021, will accelerate development. It permits numerous application scenarios without the need for a safety driver.

And they are also agreed on the fact that the next step to be conquered, Level 3 (highly automated driving), is a milestone. With Level 3, the driver not only hands over control of the steering wheel to the software, but the vehicle also takes 100 per cent of the responsibility in these defined framework conditions. The technical challenge in this is, as Woopen explains, how the system can give back driving responsibility for the vehicle to the human in an emergency. It’s a huge step. And to be able to take that step, Audi is undertaking further testing on how the driving algorithms react in the virtual representation of Ingolstadt.

Benedikt Sperling-Zikesch, Sales Director People Transport EMEA & Managing Director DACH, Easymile
Benedikt Sperling-Zikesch, Sales Director People Transport EMEA & Managing Director DACH, Easymile

‘We note that people are a lot less sceptical once they have experienced an autonomous vehicle for the first time. This underpins the importance of real laboratories that make the technology tangible. But there is still a lot of work to do on the technology too. Above all, the reliability of perception in such vehicles must be improved. Various environmental and traffic conditions pose a hurdle. At the moment, adverse weather conditions such as heavy snowfall, rain or fog still present severe limitations for highly automated mobility services. We are carrying out research into this together with strong technology partners.’

Christian Ruland, Vice President Operations Autonomous Mobility & Information Systems Bertrandt AG
Christian Ruland, Vice President Operations Autonomous Mobility & Information Systems Bertrandt AG

‘The key technologies for autonomous systems are very well developed for standard situations on motorways. But in city situations, the current technology is not sufficient to guarantee the required safety from the user’s point of view. We need technological, affordable, secure standard solutions as well as deep technical software expertise from the vehicle to the backend for individual solutions.’

Marc Rother, Global Director of Sales, Magna Electronics
Marc Rother, Global Director of Sales, Magna Electronics

‘The technical effort required for a Level 3 system is significantly higher than for a pure Level 2 assistance system. However, to ensure market penetration, new sales concepts such as features on demand or subscription fees will also increasingly come into play. Acceptance must be considered in terms of the cost-benefit aspect. If the time gained while the vehicle is driving autonomously can be used sensibly, and in the best case even in a way that adds value, then the costs for the system will be accepted.’

Dr. Christian Hort, Leiter Automotive und Manufacturing, T-Systems
Dr. Christian Hort, Leiter Automotive und Manufacturing, T-Systems

‘5G connectivity in conjunction with edge computing will without doubt play a central role in automated driving. For example, in the high-performance processing of the high volume of data from sensor and environmental data for the further development of AI or image recognition or communication with other vehicles or the infrastructure (V2X). These technologies are also central to expanding the system of giving vehicles software updates over the air (OTA). However, it is also clear that autonomous driving must also be able to function without 5G coverage as we cannot assume permanent network coverage.’


  • Future Mobility