Proceedings of the 12th International Scientific Conference on Mobility and Transport

“Mobility Innovations for Growing MegaCities”—was the topic of mobil.TUM 2022, International Scientific Conference on Mobility and Transport. Based on a historical review on the developments in transportation engineering, we wish to contribute to the ongoing transformation processes in this field. More than 60 peers joined this event held online in April 2022.

Traffic congestion might be partly solved by using autonomously driving vehicles which are expected to enter the market at a significant rate within the next years (Kaltenhäuser et al. in Transp Res Part A: Policy Pract 132:882–910, 2020, [1]; Bansal and Kockelman KM in Transp Res Part A: Policy Pract 95:49–63, 2017, [2]; Nieuwenhuijsen et al. in Transp Res Part C: Emerg Techno 86:300–327, 2018, [3]). Several studies have been undertaken to examine the impact of autonomous vehicles (AVs) on road traffic. Also, autonomous vehicles and connected autonomous vehicles (CAVs) have been simulated in the literature with different operational parameters, leading to different results. Hence, in our study we examine how different parameters for the operation of AVs and CAVs influence urban traffic in the case of Munich, Germany. Furthermore, the impact of different percentages of AVs and CAVs on urban traffic is studied. For this, the traffic will be studied for the whole city, as well as for certain travel routes, e.g. in the main travel direction (into the city in the morning), in opposite direction or along the highway surrounding Munich. Last but not least, future scenarios with an enhanced travel behaviour will be studied. The results show that the headway and reaction times of the vehicles have the largest impact on urban traffic. Here, vehicles with large reaction times have a negative impact on urban traffic while short reaction times have a positive one. The results can be used to configure future AVs such that they reduce congestions and optimize urban traffic flow.

The intelligent transportation systems (ITS) are part of possible solutions to the problems in transportation. Current systems generate digital twins of traffic participants. The traffic can be interpreted, and control signals can be sent to vehicles. Malfunctions could have disastrous consequences. Therefore, we present a self-diagnosis functionality for ITS which enhancing robustness against component failures. First, we identified sources of failures. Then, we compared existing failure detection approaches in use case of ITS. Based on this, we developed the methods Heartbeat, Sensor Metadata Checking, Process Pipeline Checking and Measurement Point Cross-Checking. To react to malfunctions, we introduce a remediation component. For testing, we used the real environment test bed Providentia++. The unique setup enables novel approaches for enhancement of robustness. In particular, Measurement Point Cross-Checking was tailored to our unique sensor setup. During the experiments, we verified the effectiveness of our methods. In future work, we suggest more plausibility checks.

As cities grow larger, they often struggle in finding sustainable and liveable mobility solutions to accommodate this growth. Many alternative modes of transport—such as public transport, carsharing systems, bikesharing systems—exist next to private car travel. The effects of expanding those alternatives are often challenging to model. This is in particular the case for small and medium sized cities, which often use straightforward and easy-to-use four-step traffic models. The alternative modes of transport could be modelled using extensive agent-based traffic models. However, these are expensive to make and require a lot of data and expertise. In the context of the EU H2020 project “MOMENTUM”, we developed an intermediate modelling approach that aims to reconcile the user-friendliness of four-step traffic models with the predictive power of agent-based models to investigate the effects of alternative modes of transport. In this paper, we demonstrate the modelling of a policy plan—away from private transport towards durable modes of transport such as shared mobility and public transport—in the city of Leuven, Belgium. We focus particularly on the developed disaggregate car-ownership model, induced demand model, and link-level emission model. It was found that an improved carsharing supply can significantly reduce the car ownership of a city’s households. The largest reduction is seen in households that own several cars and decide they can do with one fewer. These households can use the carsharing system for the occasional trip they would make with the additional car. Moreover, policy measures for the promotion of alternative modes of transport—which might increase the travel times to reach the city for privately owned cars—were found to be able to reduce the city’s mobility-related emissions. In conclusion, we demonstrated that the developed intermediate modelling approach is versatile and applicable to the cities like Leuven, such that they can also account for new modes of transport. The developed models and concepts can help other small- and medium-sized cities to shape their mobility plans.

Personal rapid transit (PRT) systems are on-demand transport systems in which small autonomous vehicles travel on their own dedicated infrastructure. Passengers travel alone or in small groups in a vehicle without intermediate stops to their requested destination. Typically, they are considered as medium capacity systems of 5,000–20,000 passengers/h. This paper sets the focus on the optimal passenger guidance within PRT stations to maximize station capacity and in particular in the queue formation for the passengers within the PRT station. The queueing processes are modelled with a microscopic traffic simulation software. The examined scenario is a train station at which the passengers form queues before boarding to the PRT system. Two types of queues are examined: An S-shaped and a funnel-shaped queue. To determine the maximum number of the dispatched passengers of the PRT station, the percentage of the passengers alighting from the train is increased incrementally. The results of the simulation show that both queue types reach the same maximum number of dispatched passengers per hour. This indicates that there is a certain likelihood that station capacity is independent of the shape of queues within the station, but it depends rather on availability of vehicles within the station. We draw this conclusion from a visual observation of the model and a rough comparison with values from a previous study. Therefore, there is more potential to increase PRT system capacities in the optimization of the PRT vehicle-side processes in the station, rather than in the optimization of passenger-side queues in the station.

Digital mobility platforms vary with respect to their design and configurations. Depending on the specific design choices, such platforms offer different value creation and value capture mechanisms. Building on research in the field of platform and existing data-driven mobility platforms, we develop an understanding of the elements and facets in which platforms are configured and construct a morphological box with these constitutional elements. To validate and exemplify our findings we draw on existing mobility data platforms and their configuration.

The paper presents a standardized method to generate agent-based transport models. In this context, the paper depicts an approach to processing multiple public and nationally available datasets and illustrates how they can be fused to create a synthetic population of agents. We then present a methodology to supply the generated agents with individual attributes and assign a schedule of activities as well as individual mobility behavior. Subsequently, we apply the methodology to the city of Hanover, and the resulting model is compared to and evaluated with real-world data. In the course of this evaluation, it becomes apparent that the pipeline generates a sufficiently accurate model that can be used to study the traffic in the city.

The wide-ranging implementation of car-sharing has led to many positive impacts but also has been a controversial issue as it can cause congestion if it replaces public transport trips. Therefore, we are aiming to investigate the competition and cooperation between car-sharing and public transport with a focus on the car-sharing pricing policy. We consider shared autonomous vehicles (SAVs) for which a range of pricing scenarios are possible and simulate different scenarios with a given public transport networks. An important aspect of SAVs is the relocation strategy to cater future demand optimally and this is considered in our simulation. We consider distance specific pricing, as well as origin–destination specific pricing for car-sharing. We create a model of Kyoto city with the real-world subway lines and trip data and simulate choices between public transport and SAVs. The performance of the whole network is evaluated by the cost for passengers, profit for the car-sharing operator as well as the spatial distribution of car-sharing prices and its modal share. Results show that if a car-sharing service is introduced in the network, cost for passengers can be decreased. The benefit is spatially heterogenous and depends on the public transport network. A main result is that the benefit is significantly reduced if the SAV operator can pick prices for all ODs freely instead of being bound to a distance-based fare. Hence, it is necessary for city planners to introduce some pricing regulation, in order to balance and maximize the benefits for both passengers and car-sharing operators.

E-commerce demand has increased steadily over the last decades and this trend has accelerated even more since the start of the Covid-19 pandemic. This entailed that user groups such as older people who previously only shopped in-store were incited to shop online to reduce risk of infection leading some to switch to online shopping as the main shopping channel. This study analyses the long-term effects of increased online shopping and subsequent delivery demand due to the Covid-19 pandemic using an agent-based travel demand model. We analyse the simulation of two scenarios for the model area Karlsruhe, Germany: one scenario simulates the parcel delivery demand before the pandemic and the other scenario simulates the demand during the pandemic of the synthetic population. Our results show that there have been shifts in both socio-demographic characteristics of online shoppers and spatial distribution of parcel delivery demand induced by the Covid-19 pandemic. The scenario simulation based on the pandemic related data shows that not only the influence of income has shifted but also the effects of age on e-commerce activity has changed due to the pandemic. The findings are of interest to transport planners and delivery service providers as they highlight the importance of recognising that the Covid-19 pandemic not only induced a shift in socio-demographic profiles of online shoppers but that this shift also entails a change in the spatial distribution of parcel deliveries.

The mobility needs of individuals with travel-limiting disabilities has been a transportation policy priority in the United States for more than thirty years, but efforts to model the behavioral implications of disability on travel have been limited. In this research, we present a daily activity pattern choice model for multiple person type segments including an individual’s wheelchair use as an explanatory variable. The model results show a strong negative impact of wheelchair use on out-of-home travel, exceeding the impact of other variables commonly considered in such models. We then apply the estimated model within an activity-based model for the Wasatch Front region in Utah; the results suggest a shift in tour making of sufficient scale—among both wheelchair users and those in their households—to warrant further scrutiny and analysis.

In this study, the effects of inner-city consolidation centers on route distance, route duration and delivery costs are evaluated using the example of Äußere Neustadt a district in Dresden and the courier, express and parcel (CEP) sector as an example of inner-city freight transportation. In order to answer the research question, a multi-level model LOCAMM (Logistics and City Architecture Multilevel Model) is presented, which forms the input data for an agent-based traffic flow simulation in the software MATSim (the Multi-Agent Transport Simulation Toolkit) in combination with JSprit. JSprit is a toolkit to solve richt Vehicle Routing Problems and Traveling Salesman Problems and is firmly integrated in MATSim. The result of the study shows that in the present case it is possible to reduce the route distance and the delivery time by using a Micro Hub. However, depending on the chosen scenario, the delivery costs are above or below the chosen comparison scenario. Furthermore, the effects of the intermodal transshipment center Bahn-City-Portal for the mentioned parameters and the link volume are described. In this study, benefits of inner-city consolidation centers on route distance, route duration and, depending on the scenario, on transportation costs were demonstrated for this case. For the Micro Hub scenarios, an increased link volume around the Micro Hub was found, but overall a lower average relative traffic volume in the delivery district.

Searching for parking (also known as cruising) is one of the key contributors to urban congestion and consequently air pollution. Providing information about parking availability is already present in many cities, efficiently guiding drivers to vacant parking spaces, either based on past data or using technologies for real-time information. Informing drivers about the expected cruising time at their destination is an additional pathway, which can impact their departure time and mode choices, and consequently improve the overall mobility system performances. One of the main challenge of cruising time estimation lies within how cruising itself is detected. This study examines the case of detecting cruising using GPS traces from trajectories. A new parametric detection method is proposed establishing speed and acceleration/deceleration related conditions. A sensitivity analysis to the method is presented along with a comparison against existing and similar GPS-based cruising detection methods and validation against labelled data. After the cruising detection, about 800,000 GPS trips were used to estimate and validate an offline machine learning algorithm to forecast the cruising time in three different urban areas in the City of Copenhagen, Denmark, with clear distinct parking conditions. Neighborhoods were divided into spatial cells for which hourly cruising times were estimated. Feed-Forward Neural Network (FFNN) and eXtreme Gradient Boosting (XGBoost) architectures were tested as machine learning algorithms and outperformed a simple moving average (RMSE gains from 62.01 to 52.57 s). The present study paves the ground for the exploration of large datasets with GPS trajectories in urban areas for tackling the lack of information on parking search. Despite the improved overall prediction power, the potential errors from the cruising detection method, lack of data needed to capture patterns when cruising time is high or the existence of many missing values due to aggregation of data could be the reason for the observed algorithm’s inability to predict the larger values of cruising time.

This work aims to address the challenges in autonomous driving by focusing on the 3D perception of the environment using roadside LiDARs. We design a 3D object detection model that can detect traffic participants in roadside LiDARs in real-time. Our model uses an existing 3D detector as a baseline and improves its accuracy. To prove the effectiveness of our proposed modules, we train and evaluate the model on three different vehicle and infrastructure datasets. To show the domain adaptation ability of our detector, we train it on an infrastructure dataset from China and perform transfer learning on a different dataset recorded in Germany. We do several sets of experiments and ablation studies for each module in the detector that show that our model outperforms the baseline by a significant margin, while the inference speed is at 45 Hz (22 ms). We make a significant contribution with our LiDAR-based 3D detector that can be used for smart city applications to provide connected and automated vehicles with a far-reaching view. Vehicles that are connected to the roadside sensors can get information about other vehicles around the corner to improve their path and maneuver planning and to increase road traffic safety.

A clear understanding of the demand patterns, is one of the key contributors to laying a firm foundation for tourist planning. In pursuit of that aim, we estimated the number of tourists at specific areas and times in Kyoto City using regression analysis and hierarchical linear models (HLM). We first discuss how to extract the tourists’ data from a “mesh population” obtained from aggregate mobile network operational data. We then propose that a relatively small sample of GPS tracking data for a population that has been monitored over a longer time than the mesh population can be used as a surrogate. To distinguish tourists from other persons, we find that a specified threshold of visiting a certain number of tourist attractions per day is useful. We also examine the effect of months and time of days by HLM on the model fit and number of tourists. Finally, we show that the accessibility of information such as the level of the attractiveness of particular Points of Interests (POIs) measured in terms of “Google ratings”, in conjunction with the GPS records significantly contributes to a better estimation of the number of tourists at specific areas and times in Kyoto City.

The majority of shared mobility services such as ridepooling use predefined locations to pick-up and drop-off passengers, so-called virtual stops. Compared to public transport stops, they do not require any infrastructural conditions. However, little to no attention is usually paid to the importance of the local environment when implementing virtual stops in an urban network. This paper presents an evaluation system to assess the feasibility of real-world virtual stops. We focus on essential criteria concerning admissibility, accessibility and the impact on traffic flow. To apply the proposed evaluation system, we used a case study of a mid-sized city in Germany with more than 4.300 virtual stops deployed by three different distribution methods. The results show that a vehicle may legally stop at 87% of these virtual stop locations and that 38% provide barrier-free access. In addition, at almost all stops (about 95%), the vehicle must remain on the roadway since there is no bay offered to leave the road. Regarding the distribution methods, the intersection approach shows benefits.

A significant amount of road traffic emissions are caused by traffic jams and stops in particular in front of (signalized) intersections. Reliable signal timing estimation methods could be the key to minimizing unnecessary braking and accelerating in front of traffic lights. In this paper, we present comparative experiments of machine learning algorithms to predict the switching times of traffic actuated signals considering classification as well as regression. Best results for the prediction of traffic actuated signals were obtained by methods based on decision trees. We came to the conclusion that extreme gradient boosting (XGBoost) shows the best performance for the two traffic signal systems under consideration. Here the most poorly performing methods were SARIMAX and neural networks (RNN, LSTM, GRU, and MLP).