Ideas for the mobility of the future

How we will travel tomorrow is being simulated and calculated with the help of supercomputers at the Department of Traffic Engineering at the Technical University of Munich (TUM).


The digitalised traffic of the future is being researched in the north of Munich. The results
improve simulations and models. Graphic: M. Markreiter, Chair of Traffic Engineering/TUM

Those who rely on public transport in Munich are often faced with disappointment: The app of the Munich Transport Company (MVG) is increasingly reporting bus cancellations. In the late summer 2023 the company is currently short of at least 50 drivers. On ten bus routes, the service has been reduced. And as in Munich, drivers are in demand in Berlin, Frankfurt and Hamburg: "We are experiencing an extreme shortage of skilled workers in the transport sector," says Dr. Klaus Bogenberger, Professor of Transport Engineering at the Technical University of Munich (TUM). "There is no way around the automation of transport." Autonomous vehicles can help alleviate this shortage and make urban mobility more modern, quieter and safer.

Using future scenarios to improve the present

Bogenberger and his team are preparing for this future with the help of the Leibniz Supercomputing Centre’s (LRZ) supercomputers. They are simulating scenarios such as robo-shuttles in Munich; how pedestrians, cyclists and autonomous vehicles will interact together on the street;  how potential passengers could use self-driving shuttles. For the theory of traffic flow, which is Bogenberger's area of expertise, the question of how traffic can become more fluid and safer is also exciting: "We solve optimisation problems. For this we need fast algorithms and calculation methods, as well as the computing power of supercomputers," explains Bogenberger. "It is possible to simulate traffic and entire conurbations, but a key point is to make the simulators needed for this more and more realistic and better."

While research disciplines usually also develop models and simulations to understand system and processes and to produce forecasts, the researchers at the Department of Transport Engineering do it the other way round: they model future scenarios, test them with the help of experiments in everyday traffic and use the real-life data to improve the models and simulations. The fact that current problems are still being solved in the process is of practical importance: "We can't know how robo-cars will be used, because apart from a few test vehicles, there aren’t any yet. Since we cannot test everything in real life, we need simulation and many different models," says the researcher. "The transport sector fascinates science because it affects the whole of society, every one of us is mobile in some way every day. Transport is a real system that politicians and administrators want to influence".

Finally, Bogenberger's team simulated ride-pooling, the on-demand transport service in which several people with different destinations share an autonomously driving minibus. The scenarios varied in fleet and vehicle size and number of passengers and routes used. They included optimisation challenges similar to the travelling salesman problem, i.e. a calculation that becomes more complex and larger with each modification. The simulations showed how the service could be organised, which routes would be heavily used and when the vehicles would need to be recharged: "We calculate very large scenarios, with four or six people per vehicle, with fleets of 3,000, 5,000 and more cars, if 10, 20 or even 30 percent of the population participate," says Bogenberger. He points out one lesson that was learned: "To be used, ride-pooling needs to be of high quality and available around the clock, but that is a complex optimisation problem that needs to be solved in real time because customers want to be transported immediately." Another lesson: If self-driving and controlled vehicles are ever to roll along a road, traffic would have to be calmed by a speed limit of 30 kilometres per hour. Otherwise, the autonomous vehicles would be overwhelmed. However, these results have nowadays to be confirmed by surveys or tests: "We assume that everyone reacts rationally to ride-pooling," Bogenberger justifies the reality check. "But what if there is a lack of trust in a car without a driver?"


EDGAR is an autonomously driving shuttle and research project, for which
the Department of Automotive Engineering has converted a minibus.

Photo: A. Heddergott/TUM


The use of autonomous shuttles in Munich is simulated
with the own open source software FleetPy
Graphic: TUM/ Chair of Traffic Engineering

Unique test field

New mobility services such as ride-pooling are the subject of the projects Minga (Munich's automated local transport system involving ride-pooling, solo buses and bus platoons) TEMPUS and the MCube cluster for the future of mobility in Munich, in which the Munich Transport Authority (MVG) and Stadtwerke München (Munich utility company) are involved and which are funded by the Federal Ministry of Digital Affairs and Transport (BMDV). Buses that drive autonomously or automatically follow a human-driven bus are also being tested. For TEMPUS, the test bed for automated and networked mobility in Munich and the surrounding area, autonomous cars from BMW and automated buses from Ebusco are time and again on the road in the north of Munich, between the A9 and A99 motorways and the Olympiazentrum and Unterschleißheim. How cameras, drones and contact loops control roads and traffic in the test area was demonstrated at the IAA Mobility 2023 International Motor Show. During the field test, selected cars and drivers received also real-time data on traffic jams and accidents via 4G and 5G networks, and traffic lights also used the data to regulate the flow of traffic: "Such a test field consisting of all road categories is so far unique in the world so far; it is not only about automated traffic in the city, the preparation of the infrastructure and new transport concepts," says Bogenberger. "Another goal is the standardisation of data flows." Information from smart traffic lights, cameras, test vehicles and service apps will flow together and be automatically harmonised in an automated way to make it usable. This information could then also be used to create digital twins of mobility services or of roads, which in turn could be used by control centres to manage events on rail and road more intelligently.

Freely available data in demand

This is still a vision of the future. However, both simulated scenarios and real-world experiments are laying the groundwork: "Every device manufacturer, mobile phone provider or car manufacturer stores motion data worldwide. The big dilemma for researchers is that they have no access to this data," says Bogenberger. That's why they rely on cooperation with companies at the chair, buy data - or collect it themselves. Measures in Germany such as the €9 or €49 tickets not only raise specific research questions, they also bring with them new technical possibilities for collecting data that make future scenarios more detailed, more realistic and better. Since 2022, for example, the transport engineers have been using an app to continuously collect the movement data of more than 1,000 participants, who are occasionally asked about tickets and driving behaviour. "From this we derive a great deal of knowledge about the mobility behaviour of Munich residents," says Bogenberger.

Of course, these and other data sets have since been analysed using artificial intelligence (AI) methods. "Traffic constantly produces recurring patterns, such as traffic jams during commuter or holiday traffic. AI can identify these recurring effects," Bogenberger reports. "But it doesn’t yet make sense to train neural networks with the results of our future scenarios or simulation data from large fleets. We still need a few more empirical experiments and data enrichment, because today the AI would be learning nonsense." Traffic technology is based on a vision of the future in order to address and solve questions of the present. And that's how it should be, according to Bogenberger: "After all, the task of science is to develop and offer solutions at an early stage that politicians or society at large can accept or reject." (vs/sschl)


Prof. Klaus Bogenberger, Chair of Transport Engineering, TUM