Hear the experts: Transportation modelling trends in 2021

In teaching and research especially, an exchange of ideas and discussions with colleagues is essential for personal development and the success of one’s own research work. In February 2021, the PTV Academic Forum took place for the first time. A platform, created for tutors and researchers in the higher education landscape, with the purpose of discussing current trends in transportation modelling and learning from each other. Particularly exciting this year was the break-out session with renowned experts on the topic of “Mobility trends: Innovation in research”.

We have summarized the key statements on the 2021 mobility trends for you. Have fun!

Trend “The bicycle in transportation modelling”

Elisabeth Lerch, University of Applied Science Frankfurt

At the Frankfurt University of Applied Sciences, our main interest within the framework of the new endowed professorship for bicycle transportation lies in identifying how we understand and structure our teaching assignment. Our most important task at present is to investigate which research fields already exist on the subject of “bicycle transportation” and where there is the greatest need. We can then define in which direction our own work should be going.
To give you a small insight into our approach: We want to know, for example, how do cyclists behave / how do pedestrians behave? How do they behave when they have to share the same space? Unfortunately, there isn’t sufficient data on active mobility forms at the moment. And due to COVID-19, we are not able to carry out inner-city surveys right now because the situation is out of the ordinary. Therefore, the knowledge we would gain would be unusable for our purposes. Nevertheless, it will be exciting to see if the trends caused by COVID-19, i.e. more people working from home or an increasing share of bicycle transportation, will remain when the pandemic is over. In the field of leisure usage however, plausible bicycle transportation data can be surveyed because this is where bicycle usage is booming. So this is where our focus currently lies.

Dr. Klaus Nökel, PTV Group

In my opinion, the modeling of bicycle transportation comes under the central issue of “The redistribution of road space”. We need models that are sensitive on quite different levels. For example, does a junction remain “walkable” when it is redesigned for bicycle transportation? This can be depicted very clearly using PTV Vissim, as the software takes all road users and their behavior parameters into account.

Prof. Dr. Christoph Walther, PTV Group

We at PTV have been researching bicycle traffic for many years now. Even over 10 years ago, we were already developing a macroeconomic assessment method for bicycle infrastructures. And we are currently in the scientific expert team supporting the development of the National Cycling Plan (NCP) for the Federal Ministry of Transport and Digital Infrastructure (BMVI). I personally find the thought interesting of one day investigating the development of bicycle transportation to see if there could be a break-even point in demand. A point from which the positive effect would decline, so that it would possibly no longer do the traffic flow any good if even more citizens were to switch to their bicycles. In Copenhagen, I experienced not being able to cross over to the bus stop as pedestrian because so many cyclists were on the road. A strong majority of cyclists does not necessarily lead to an optimum utilization of the road space.

Prof. Dr. Matthias Richter, University of Applied Sciences Zwickau

We recently carried out an empirical study on the topic of “Causes and effects of conflicts between cyclists and pedestrians” in cooperation with the Cracow University of Technology. My colleague Marek Bauer and I presented the results at the “13th International Road Safety Conference, GAMBIT 2020”. The study analyzed the behavior of cyclists and pedestrians and their interaction depending on various factors, such as the infrastructure available for bicycle transportation. One interesting observation we made was that the share of pedestrians who are responsible for conflicts is constantly rising and that respective measures need to be and can be made in order to make the coming together of various road users more conflict-free. In particular in terms of the safety of the two groups on pedestrian and cycle paths and on shared spaces. Unfortunately, many pedestrians and cyclists encounter the other road user negatively and sometimes also inconsiderately. Experience shows us: someone who is a rowdy in their car isn’t an angel on their bicycle or as pedestrian either.

Michael Cik, Graz University of Technology

A massive mobility trend in Austria is bicycling. At Graz University of Technology, we are concerned with how we as mobility experts can provide a good and correct representation of bicycle traffic in a mobility model in the future. The interest in active transportation modes in cities and local authorities hasn’t just been very high since the pandemic. In research, we haven’t only got an eye on the characteristics of the transportation mode (as input providers in the model technique), but also the data base. How do I get the information I require? Do I use GPS data? Do I set up counting stations or do I purchase mobile communications data?

Trend “Data in transportation modelling”

Dr. Klaus Nökel, PTV Group

It is a study of empiricism: Despite the fact that COVID-19 is a great burden for industry, society and economy, the situation is still very interesting from the point of view of data science. Scientists are collecting data like crazy in order to dissect and analyze this global “major experiment” as precisely as possible in terms of behavior changes. At the moment, we can only speculate on how many of these behavior changes will permanently establish themselves as mobility trends. It is a fact though that several of the characteristics have only been accelerated by COVID-19 and are therefore sure to permanently manifest themselves as new behavior patterns. However, we cannot yet say which ones those will be and so they cannot be included in the model either.

Michael Cik, Graz University of Technology

Getting and preparing relevant data is already 30–35% of the work in advance. We currently have a project in Graz which is investigating active mobility. The focus is on users of basic scooters (not electric scooters) and cyclists. The study begins with the question: How do I gain data records for the research subject with which I can generate results? Getting comprehensive information on certain reference dates involves a great amount of time and expense. There’s not one recipe for all. One thing which we have definitely included in our teaching assignment is that we train the students to be able to deal with statistical analyses and data processing.

Trend “Artificial intelligence in transportation modelling”

Dr. Klaus Nökel, PTV Group

Transportation modeling can no longer be imagined without artificial intelligence. It is to be seen as extension to conventional methods of transportation modeling or even, to some extent, already a replacement for it. In my opinion, we, the transportation experts at PTV, have to make a decision: In future, will we be modeling mobility decisions in our products such as the choice of transportation mode based on random forests or neuronal networks instead of the logit models that are currently everywhere in transportation models?
Transportation modelling in real-time has great potential for the short-term forecasting of traffic conditions, the core task in traffic management. Artificial intelligence could be used to define traffic conditions using video images of a road section or junction and, as a result, to be able to derive simple control actions. Thus, AI-based methods could keep the background noise in every-day standard operations under control, whereas in the case of a major incident on the highway, it’s still humans who are needed.
If we want to advance this development, there’s no way around high-quality training data records for artificial intelligence. An exciting question is whether transportation models can be used to synthesize such training data.

Andreas Stadler, PTV Group

Data was, is, and will be increasingly important for transportation modeling. We need to find a way to continuously survey data and then to use it to extend or improve the models. Ultimately, it is a question of maintaining a living model, regardless of whether I mean an urban model or a small-scale Vissim model. The data base is, in my opinion, more important than the methodology. Whether I use new AI or machine learning methods for it, or an old statistical procedure: they all require data. The difference between the procedure results is smaller than the difference which would have occurred if the data quality had diverged. The really interesting questions, for example: How can shared mobility be modeled?, How will things continue with COVID-19?, How do people react to a change in bicycle transportation? These can, up to a certain point, be meaningfully modeled in theory. We are just struggling with the missing data base.

Prof. Dr. Matthias Richter, University of Applied Sciences Zwickau

On the use of AI methods in transportation management
As of late, Zwickau University of Applied Sciences has carried the name “University for Mobility”. As such, we, in a similar way as described for Frankfurt University of Applied Sciences, are still in the process of self-discovery and are currently defining the focus this means for our work. “Mobility” can be interpreted in many different ways and comprises countless aspects from fields such as vehicle and production, energy and infrastructure, digitalization, but also health and sustainability. In the field of transportation management, for example, we are carrying out tests on the use of AI-based models. There is also a lot of pure craftsmanship involved in this: Putting up a camera or flying over traffic using drones, identifying vehicles and determining parameters. I believe that in this point in particular, there is great potential for research with real-time data. With comparably simple methods, great effects can be achieved in this way which, for example, with regard to the use of autonomous vehicles (and also their co-existence with classic vehicles) open up unforeseen possibilities.

Michael Cik, Graz University of Technology

It gets exciting when I can work out the activities of people from the models. In time, we will become increasingly better at mapping activities using artificial intelligence. Regardless of whether these are rule-based approaches (random forest, neuronal networks) or probabilistic model approaches – the models become increasingly clever with both procedures, but can only function if we generate good training and test data records in order to train the AI. My dream would be an annotated database of a suitably large population group. The data records would have to contain details on the trajectory, the activities and the daily schedule of each individual person. The model is trained on the basis of this training data and gets increasingly clever over time. With a second data record I then validate the model. In my opinion, it would be important to invest research funds into these data surveys, because these findings should also be included in future household surveys, which of course should be improved and digitalized over time.

Trend “Transportation modelling in an inter-disciplinary context”

Michael Cik, Graz University of Technology

For me, the greatest insight in recent times was to see how many specialist fields really do interact and can work together on the basis of a demand model, which you might not normally think of. It is possible to use the empirical facts, which you inherit, and the demand model, that you create, for so many different purposes. Whether that’s air, noise or environment or, as is currently the case, the health sector. It is interesting to see what impact mobility and demand have on the other research fields.
For example, when I compare the mobility curve in Austria for March 2020 and that of the second and third lockdown with the economic development curve from this period and the development in the epidemiological area, clear correlations can be seen.
However, in a conversation with the Institute for Economic Research we learned that the calculation results from the transportation model are needed very quickly for further processing in order to be able to make economic forecasts, for example.

Dr. Klaus Nökel, PTV Group

If COVID-19 has shown us one thing, it is that we have to expand our field of activity in transportation modeling. We are so used to consulting transportation modeling for infrastructure planning alone, that we are shamefully neglecting further possible areas of application beyond this. Infrastructure planning doesn’t play a role in the case of the pandemic. Transportation models were and are used here to determine contact networks and risks of infection and to find out what effects closures could have on the economy. The interdisciplinary possibilities which have now emerged are more than exciting.

Prof. Dr. Christoph Walther, PTV Group

In 2019, the European Union started the GAIA-X initiative with the aim of reducing the dependency on American and Chinese IT providers and data-driven, market-dominating platforms. GAIA-X provides an organizational and legal framework for a respective cloud architecture. The Mobility data room initiated by the German Federal Ministry of Transport and Digital Infrastructure (BMVI) could become an application example for GAIA-X in the mobility field. As user, the opportunity of such data rooms lies in working with data records from different providers and thus to generate better matrices than demand matrices on the basis of only one provider. This could also revolutionize the German Federal Transport Plan if planners don’t have to put disproportionate effort into updating the matrices every 5-10 years, but instead continuously update them. Many European countries and companies around the world are already supporting the project.

Trend “Transportation modelling in general”

Dr. Klaus Nökel, PTV Group

We have to convince the cities and local authorities of the benefits of transportation modeling. We therefore need to make the use of models more affordable and attractive for smaller municipalities. Smaller towns often don’t have the financial and human resources to create a model and to adequately maintain it. Here, actions are currently taken to reduce the time and costs involved in the construction of models. I also think we have to orientate ourselves more around designing our products in such a way that they are more easily understandable, so that colleagues, who are not trained transportation specialists and who only deal with transportation occasionally, can find a way into the topic.
Mobility experts should also move away from seeing transportation models as a big dusty folder which is taken out of the drawer to work with once a year. A model should be used more for monitoring purposes: every week, every day, in order to check if the model corresponds to the reality on the other side of the window. In this way, changes in the transportation process could be recognized early and the cause identified so that data-driven recommendations for action can be voiced.

Michael Cik, Graz University of Technology

Until now we have always worked in a sterile laboratory situation but, since March last year, this artificial environment has become a reality, thanks to COVID-19. We are now concerned with other questions which we are investigating in lead projects. How will mobility continue to develop? Which changes in mobility behavior can we observe? I haven’t heard from anyone yet that they have found the ultimate formula on how mobility will develop after coronavirus. This is, I believe, an exciting research subject, not only for the field of transportation modeling. For one thing, we are trying to integrate this data-driven model, in which we work with mobile communications data, into agent-based models. We are also working very closely with colleagues in Switzerland. I have always found it very exciting what Mr Nagel (editor’s note: Prof. Kai Nagel, TU Berlin) with his team and the RKI are doing on the subject.

Prof. Dr. Christoph Walther, PTV Group

In countries with a highly developed and dense transportation network, a necessary trend is to prioritize the preservation of the infrastructure over further expansion. With the Federal Transport Plan 2030, we may have managed the shift from new construction to preservation, however an optimization procedure for this preservation process, i.e. the prioritization of preservation measures on a limited budget is missing. The data base for preservation planning can be increasingly refined: In principle, cars are in a position to record uneven surfaces when driving on the streets. This data provides a good indication of the current state of the road surface. Sensors, so-called intelligent infrastructures, can provide information on changes to the structure. Data collecting and data mining could thus revolutionize preservation planning.
Climate protection eclipses everything. Our government has committed itself to climate protection goals (e.g. -42% CO2in the transportation sector), the implementation of which is very ambitious. Initial measures won’t become effective until 2–3 years have passed. That means that the linear reduction function is vastly predestined to fail in the first few years. On the one hand we could manage to just about make it at first – “thanks” to COVID-19 – but the buffer will probably be used up fast. If the goals are not achieved, the law states that emergency measures will have to come into force. We will undoubtedly encounter interesting planning challenges at this point.