Available Masters' Theses

This master's thesis aims to develop a methodology for deriving historical and geographical pollution data of a city using data from monitoring stations. The following aspects shall be considered: 1. Systematic removal of weather and event influences to eliminate distorted data. 2. Focus on pollution data associated with traffic influences. 3. Mapping of monitoring stations onto the urban traffic network. Methodologically, the reliability of the results will be assessed, and various approaches will be discussed to verify and potentially adjust them. The specific application example will be the temporal evolution of pollution in Paris from 2015 to 2023. Proficiency in Python is advantageous and recommended for implementing the methodology.

This thesis aims to analyse a dataset obtained from Munich public transport operator MVG, which includes data recorded via the various devices installed on public transport vehicles. The primary objective is to extract the speed profiles of these vehicles and cluster them to identify patterns in routes and temporal variations. Also, the study will compare the speed profiles of public transport vehicles with loop detector data to provide comprehensive insights into urban traffic dynamics. [Cosupervisor: Allister Loder, allister.loder@tum.de]

This thesis focuses on utilising a dataset obtained from Munich public transport operator MVG, which includes data recorded via the various devices installed on public transport vehicles. The objective is to develop a deep-learning method for imputing passenger counts in the dataset, extending from a subset of trips to all trips. Also, identify service potential across the most overcrowded routes, thereby providing insights into optimising public transport services within the city. [Cosupervisor: Allister Loder, allister.loder@tum.de]

The master's thesis aims to investigate how political measures can be best quantified. Various political approaches exist to curb traffic, including parking pricing, reducing parking spaces, and the concept of the "15-minute city." To determine the effectiveness of these measures, they must first be quantified. This work aims to develop a methodology to effectively quantify political measures. The reliability of the results will be examined, and various methods will be discussed to validate and potentially adjust them. This includes verifying data integrity and applying validation techniques to ensure the reliability of the results. Paris serves as a case study, utilizing official data from Paris Open Data in conjunction with information from Google StreetView and OpenStreetMap. Proficiency in Python is advantageous.

For children and families, different conditions apply when it comes to choosing a suitable bike route. For example, it is allowed to ride on the sidewalk with small children. The work is about identifying the relevant factors that significantly influence the choice of bicycle routes for children.

This thesis focuses on recognizing the intentions of extra vulnerable road users (eVRU), specifically wheelchair users. The analysis relies on camera and Lidar data, utilizing both conventional algorithms and deep learning approaches.

This thesis focuses on recognizing the intentions of vulnerable road users (VRU), specifically pedestrians and cyclists. The analysis relies on camera and Lidar data, utilizing both conventional algorithms and deep learning approaches.

With the help of augmented reality, a driver will interact with virtual road users and objects while driving on the test field in order to combine the advantages of hazard-free investigation of traffic scenarios with the high validity of test field studies.

Mobility service operators can influence demand through dynamic pricing. After a summary of existing methods, different algorithms will be implemented and evaluated.

The aim of this thesis is the extension of a concept for the quantitative, data-based determination of the stress level for pedestrians based on static, dynamic and individual influencing factors. To this end, weak points are to be identified and improvements derived and implemented. In a second part of the thesis, the final concept will be validated and calibrated by means of a survey.

The aim of the thesis is to study the routing of a fleet of vehicles (with big batteries) that can charge other on-demand ride-hailing or ride-pooling vehicles (with smaller batteries). The thesis will first research the available methods for the routing of on-demand charging vehicles, develop a new routing scheme and then evaluate their efficiency in an agent-based simulation framework, called FleetPy.

The aim of the thesis is to study the routing of a fleet of vehicles (with big batteries) that can charge other privately owned electric vehicles (with smaller batteries). The thesis will first research the available methods for the routing of on-demand charging vehicles, develop a new routing scheme and then evaluate their efficiency in an agent-based simulation framework, called FleetPy.

The thesis should evaluate and improve a draft concept of evaluation for Urban Air Mobility (UAM) systems. These include a review of proposed KPIs and indicators (based on DIN EN 13816). Next, the improved concept of evaluation should be applied to evaluate a simulated UAM service (the model will be provided). Recommendations for the industry and traffic planners should be derived.

The thesis should propose a method for routing eVTOL aircraft in urban areas. For each direct route, an alternative set of waypoints should be returned. A concept of multi-layer non-fly zones should be developed (e.g., layer 1: population density, layer 2: military areas, etc.). The city of Munich should be used as a case study area. An example UAM network will be provided.

This thesis explores integrating autonomous vehicle navigation through urban construction sites into existing path planning algorithms, focusing on adapting for safe passage. The evaluation of the adapted planning behavior is conducted and assessed within a simulator, forming a vital part of the research.

This thesis focuses on integrating dynamic stops into the route planning of automated shuttle buses, aiming to efficiently adapt existing planning systems for urban transport networks. The evaluation of the planning behavior is conducted and assessed in a simulator, constituting a central part of the study.

Network clustering is important in transportation engineering to analyze the dynamics at an aggregated level. However, road classes are seldom included because of the increasing computational costs. In this thesis, an efficient clustering method is expected to include road classes as well. Starting from loop detectors' data, the student shall cluster road sections (edges) and seperate different road classes. These clusters are further used for traffic state estimation.

This thesis focuses on utilising a dataset obtained from Munich public transport operator MVG, which includes data recorded via the various devices installed on public transport vehicles. The objective is to reconstruct vehicle trajectories using GPS points. The ultimate goal is to match these trajectories to the network graph of the city. [Cosupervisor: Allister Loder, allister.loder@tum.de]

The master's thesis aims to investigate and compare quantitative methods for supplementing detector data. Detector data is often incomplete, which compromises its usability. Within the machine learning community, there are now effective procedures for approximating missing data, known as "imputing missing values". The primary objective of the quantitative analysis is to test and compare various machine learning methods using a geodataset. The geodataset pertains to Paris and documents traffic behavior since 2015. Since 2015, Paris has undergone significant network changes, which are reflected in traffic patterns. These impacts will be examined through the analysis results. Prerequisites for the analysis include basic knowledge of Python. Ideally, the student should have advanced Python skills and experience with machine learning libraries such as scikit-learn.

In ride-pooling services, customers book a trip with the provider via an app and the trips are dynamically integrated into the route planning of fleet vehicles. No-shows of a booked trip can result in costs for the operator as well as for other customers. The goal of this work is to integrate no-shows into the existing simulation environment "FleetPy" and to simulate the effects on the overall system.

In the future, automated shuttles should be able to stop flexibly almost anywhere in the transportation network. However, depending on the vehicle’s passenger (e.g., elderly persons, wheelchair users, ...), the existing infrastructure must also be included in the selection of a flexible boarding point. Not only accessibility, but also traffic safety play a major role within this selection. In this work, a tool for automated classification and evaluation of existing infrastructure will be developed.

Calibrated traffic simulations play a crucial role in traffic planning. However, traffic data often comes with spatial limitations due to administrative boundaries, posing challenges for traditional model calibration approaches, especially in regions with cross-border traffic. This thesis aims to develop a process to create and calibrate traffic simulations using Floating Car Data (FCD) to improve modelling accuracy and reliability by relying on a cross-border traffic data source. [Cosupervisor: Gabriel Tilg, gabriel.tilg@transcality.com]

In this thesis, the aim is to examine the impacts of real-world, multi-day road closures by utilising microscopic traffic simulation. The focus will be on developing traffic assignment and rerouting strategies, which will then be validated by utilising available loop detector data. Through comprehensive scenario evaluations, insights into the resilience of traffic networks are concluded. [First supervisor: Joel Brodersen, joel.brodersen@audi.de]

Traffic scenarios in road traffic or on experemintell test fields can be recorded by various sensor technology, such as cameras. Besides detection algorithms, image segmentation algorithms can detect and classify the exact outlines and areas of objects. In this thesis, a methodology will be developed to apply and analyze the "Segment Anything" algorithm from Meta (formerly Facebook) to traffic data.

This thesis explores Facial Emotion Recognition through Convolutional Neural Networks, aiming to advance human-computer interaction by accurately identifying human emotions from facial expressions.

This project focuses on integrating e-scooter simulation within Unity and VR environments, employing an actual electric scooter for immersive and realistic user experiences. It aims to bridge the gap between virtual simulations and real-world scooter maneuvering, enhancing training and entertainment applications.

Die mikroskopischen Verkehrssimulatoren basieren auf mikroskopischen Fahrzeugmodellen, d. h. auf Modellen für das Verfolgen von Fahrzeugen und das Wechseln der Fahrspur. Ziel dieser Arbeit ist es, das Verhalten der bestehenden Modelle mit hochauflösenden HighD-Datensätzen zu evaluieren und zu bewerten. Da wir davon ausgehen, dass solche Modelle das Fahrverhalten auf deutschen Autobahnen nicht vollständig abbilden, ist ein weiterer Kalibrierungsprozess erforderlich.

Non-motorized road users often behave contrary to traffic regulations. Reasons for this can be time savings and convenience, but also interactions with motor vehicles. In this MA, this behavior is investigated based on drone videos along the Rheinstraße. For this purpose, such scenarios will first be found, then patterns will be recognized in the trajectories (also available) and a rule base will be derived and implemented.

During the automated detection of objects in drone videos, errors occur in the trajectories, especially of cyclists and pedestrians, for example due to the large number of road users in a small space or shadowing by trees and buildings. Using the original videos, the extracted trajectories of VRU are to be checked for plausibility in a provided data set and then filters for the detected errors are to be developed and implemented.

The goal of this thesis is to perform a comprehensive analysis of characteristics that are considered when designing and developing human driver models for mixed traffic with CAVs in order to have an overview of existing models, what differentiates these models and its impact on mixed traffic

In rural areas close to cities, private vehicles dominate commuter traffic, but ridesharing offers a promising solution to lower CO2 emissions, reduce traffic and ease parking. However, flexible working arrangements pose a challenge. This thesis examines commuter data to measure flexibility and analyses factors such as reduced flexibility and increased demand on the efficiency of ridesharing. Join us in examining commuter behaviour and operational efficiency for a sustainable transport future.

Research question: how to fuse open-sourced data based on their characteristcs and quantify their importances? The thesis suggests the student to: 1. Search existing data fusion techniques 2. Analyze open-sourced public transportation data in Hamburg, Germany 3. Combine different types of data sources to improve the traffic analysis and explain the reasons to include certain types of data (optional) 4. Building simulation and calibrate the model

Research question: how to choose a right map matching algorithm balancing accuracy and computational time? By comparing the performances of different map-matching algorithms, the student is expected to find a balance in the accuracy of map-matching of GPS trajectory data to network graph for a dynamic and static use case within a reasonable computational time.

The MobilityCoin System - a novel traffic management system that promotes sustainable transport based on mobility budgets. A survey was conducted to find out how users of the system use their MobilityCoins. In this master's thesis, the results will be derived into a 'spending choice model' by applying a multinomial logit model (MNL) and/or by training a neural network.

The MobilityCoin System - a novel traffic management system that promotes sustainable transport based on mobility budgets. A survey was conducted to find out how users choose their mode of transport while using the MobilityCoin system. In this master's thesis, the results will be derived into a 'mode choice model' by applying a multinomial logit model (MNL) and/or by training a neural network.

The increasing penetration of electric vehicles results in the need to derive strategies for charging electric vehicles. One possible optimization variable here is to reduce the energy costs for the charging event by shifting the timing of the energy demand. Data from real charging events will be evaluated and an agent-based simulation model will be developed.

In this master thesis, three goals are pursued: (1) to duplicate data and models from an existing (online) platform for (specified) offline studies, (2) to implement procedures to incorporate real-time data into testbed(s), and (3) to investigate data aggregation with respect to traffic control.

Given a dataset of mobility tracking data, where a large share of purposes is known, it is of interest to use this data to gain more insights into the travel patterns related to particular modes and to use this knowledge to impute purposes of trips where the purpose is unknown. For purpose imputation a rule-based and a ML approach should be used and potentially combined (e.g. rule-based including land-use information for work trips and a ML-model for others).

In many large-scale sporting events, such as cross-country skiing marathons, bottlenecks and overcrowding result in the formation of congestion analogous to that observed in road traffic. Using publicly available split times as stationary detector data and GPS tracks as floating car data, the goal of this thesis is to perform a detailed analysis of the congestion formation in several high-profile cross-country ski marathon events.

In transport research, population is often segmented into groups with homogeneous mobility behavior / characteristics. The idea of visualizing the activity behavior of each individual and then clustering these images according to "traditional" unsupervised ML methods has been explored for longitudinal panel data (DOI: 10.1016/j.trip.2020.100264). In this thesis, such an approach will be adapted for a large-scale dataset of mobility tracking data collected in the Mobilität.Leben project.

Past research has shown the existence of biases in stated preference (SP) data when compared to real-world mobility patterns. Using the dataset from the Mobilität.Leben project, this thesis should examine the differences between mobility patterns declared in surveys and those recorded with a mobile app. The goals are to identify potential sources of bias that impact data reliability and propose strategies to mitigate them, thus enhancing the reliability and usability of SP data.

The goal of this thesis is to further develop a new pedestrian routing algorithm for use with social force models under development at the Chair of Traffic Engineering and Control, including evaluating several variations to the algorithm formulation with the ultimate goal of producing more human-like behavior in existing models. The framework is being developed in Python.

Hard shoulder opening is a traffic measure that provides additional capacity during high traffic demand. In this topic, based on real traffic data of a German motorway section, macroscopic and microscopic traffic flow analysis will be carried out. In addition, a microsimulation model of the same network will be developed, calibrated, and used for further analysis.

Based on the classical distinction between user equilibrium and system optimum for routing decisions in a network, a variety of approaches exist. The goal of this work is to present selected methods in detail based on a literature review and to evaluate them in a suitable test environment.

Artificial intelligence methods are currently driving the development of automated driving. This thesis focuses on how the interaction between automated vehicles (and other road users (optionally)) can be optimized with the help of artificial intelligence. Existing conventions on right of way or fixed lane allocation may be challenged.

Local detectors such as induction loops determine macroscopic traffic parameters minute-by-minute. In contrast, FCD contain vehicle trajectories. The aim of the work is to estimate the penetration rate of app-based FCD compared to local sensors on different freeway sections in the greater Munich area and thus to derive an evaluation metric of the data quality.

In the greater Munich area, congestion patterns on freeways have so far largely been identified from induction loop data (over-head displays). As soon as traffic conditions turn to non-free-flowing, a traffic control system is often activated which influences the current traffic (e.g. speed reduction). The aim of this work is to investigate congestion events on uninfluenced sections and to identify patterns in the FCD.

The goal of this work is to implement a max pressure control for an urban network in python. The max pressure control is a decentralized control in which a separate control decision is made for each intersection based on the queue lengths upstream and downstream, thus keeping a high vehicle throughput in the network.

Perimeter control is a measure of urban traffic control in which the inflow into the city is reduced on purpose via traffic signal control. It is a challenge to identify suitable points in the network for perimeter control. For this purpose, the macroscopic fundamental diagram (MFD) is increasingly used in the literature. In this master thesis, a perimeter control based on the MFD shall be implemented in a microscopic traffic simulation.

The goal of this work is to develop an urban traffic signal control based on reinforcement learning (RL), an area of machine learning, and to implement it in the microscopic simulation platform SUMO. The implemented RL- control shall be benchmarked against a common rule-based traffic-actuated control.

This topic explores the SUMO functionality to allow for dynamic lane configuration for motorways. During the simulation run, the lane layout should be changed, and the vehicles should adapt their behavior based on the new lane layout of the network.

Traffic data for traffic management and control in urban areas is derived from different data sources. Every data source suffers from certain imperfections that have to be dealt with using a dedicated quality management concept. This thesis aims at developping a good QM concept for those data sources.

Traffic flow on freeways is dependent on the environmental conditions. This thesis shall analyze fundamental dependencies between environmental conditions and traffic flow parameters - especially the capacity of freeways.

This thesis aims to create a simple machine-learning model for distinguishing and identifying individual vehicle trajectories within a dataset produced from loop detectors. The approach involves utilizing loop detector data to visually represent vehicle trajectories and employing artificial intelligence to match individual trajectories with unspecified time information into these visualizations. This work will include a literature review, an implementation phase, and a model evaluation.

Th introduction of autonomous vehicles requires them to communicate with other road users - especially with pedestrians and byciclists. This group usually does not have special communication equipment available but has to rely on infrastructure based information. One way of presenting information could be near-field-projection, that allows vehicles to project information (even traffic signs or "mobile crosswalks") directly on the road. The aim of the thesis is to evaluate the potential of such systems and to their use in information concepts using volunteer studies.

Various systems for congestion mitigation are developed and in use. In this work, a comprehensive comparison of tokens, tolls and taxes shall be developed in respect to effectiveness, efficiency and equity.

Since in the lane-free environment vehicles have the freedom to choose any lateral location, it is crucial to assign them an optimal lateral location especially when they approach a diverging area like an off-ramp. Based on the available approaches for the lane-based lane choice models, this work proposes and evaluates similar approaches for lane-free traffic.

Similar to lane-based traffic, where vehicles choose a proper lane upon an intersection, in the lane-free case, the vehicle should also adjust their lateral location properly before reaching an intersection depending on their turn movement. This work develops optimal approaches for choosing lateral locations for vehicles.

Lane-free traffic allows connected and automated vehicles to move across the entire road surface in a flexible fashion. This allows for a reorganization of today's traffic control at intersections. To conduct an evaluation of the developed ideas, a self-developed simplified simulation environment (e.g. Cellular Automata) can be used.