Theses
On this page, we list theses that were submitted by members and students of the MARS Group.
Master
Evaluierung von spezialisierten und generischen Modellen zur Identifikation von städtischen Einzelbäumen mittels Transfer Learning
Author: Daniel Osterholz
Type: Master's thesis
Semester: Winter semester 2024
Description: This thesis investigates the performance of generic, easily accessible, models (Segment Anything Model (SAM)) and their adaptation to specific domains, such as tree canopy detection in urban forests. To evaluate the performance, a model specialized for this
domain (DeepForest (DF)) is compared with a finely tuned generic model.
Indication of wet and dry periods in Germany using machine learning on meteorological and remote sensing data
Author: Justin Tran
Type: Master's thesis
Semester: Summer semester 2024
Description: This thesis investigates the calculation, comparison, and forecasting of the Standardized Precipitation-Evapotranspiration Index (SPEI) using diverse data sources and methods.
Der Data Hub: ein Geoinformationssystem für reproduzierbare Datenverarbeitung, informiert durch epidemiologische Bedarfe
Author: Jonathan Ströbele
Type: Master's thesis
Semester: Winter semester 2023
Description: As part of this work, a geographic information system (data hub) was developed that integrates a variety of data types and sources and provides a decision support system for epidemiologists based on reproducible risk assessment frameworks.
Anwendung von Reinforcement Learning und Potentialfeldern für eine Multi-Agenten-Steuerung für Drohnen zur Lokalisierung von Funksignalen
Author: Katja Schöttler
Type: Master's thesis
Semester: Winter semester 2023
Description: This thesis deals with the generation of potential maps with a CNN autoencoder and reinforcement learning. For this purpose, a simulation environment developed with the MARS framework is used to explore and evaluate the generated maps. In addition to the technical setup, the results of the various network configurations tested are explained.
Bachelor
Tensorbasierte Agenten zur Ausbreitungssimulation von Infektionskrankheiten
Author: Ersan Baran
Type: Bachelor's thesis
Semester: Summer semester 2024
Description: In this bachelor thesis, the strengths of multi-agent systems and neural networks are combined to develop a novel simulation model that takes into account individual behaviors, spatial heterogeneity and dynamic interactions. This innovative approach promises more realistic and powerful simulations and thus better support for the development of effective strategies to combat infectious diseases.
Entwicklung eines Multiagentensystems für die Simulation des Positionsangriffs im Handball unter Verwendung des MARS-Frameworks
Author: Lucas Neitsch
Type: Bachelor's thesis
Semester: Summer semester 2024
Description: This bachelor thesis describes the development of a multi-agent system for the simulation of positional attacks in handball using the MARS framework, which enables coaches to test their attacking moves in specific game situations and serves as a tool to improve tactical planning and decision-making.
Automatisierte Objekterkennung und -zählung von Vögeln in großen Schwärmen mit Hilfe von Maschinellem Lernen
Author: Luise Kempa
Type: Bachelor's thesis
Semester: Winter semester 2023
Description: The thesis deals with the implementation of bird counting in flocks based on images using machine learning. An application is designed and implemented that determines the number of birds in an image. For the implementation, different existing neural networks are compared and then extended.
Modeling and Predicting Routes of Internally Displaced Persons: An Agent-Based Approach Using the MARS Framework in the Context of the Syrian Refugee Crisis
Author: Viviam Clara Ribeiro Guimaraes
Type: Bachelor's thesis
Semester: Winter semester 2023
Description: This thesis investigates the application of an agent-based model (ABM) to predict the distribution of Syrian internally displaced persons (IDPs) and assess its utility for humanitarian resource allocation. Leveraging the MARS framework, the study adapts an existing ABM designed for Syrian refugee distribution in Turkey to simulate the movement of internally displaced persons within Syria.
Identifikation Sozialer Gruppen von Fußgängern in Videos, Motion-Similarity-Clustering und Multi-Object-Tracking unter Anwendung eines CNNs zur Identifikation von Personen in Videostreams
Author: Tobias Ranfft
Type: Bachelor's thesis
Semester: Summer semester 2023
Description: Real-time recognition of grouped people in videos has recently become a relevant topic in research and for various applications, e.g. video surveillance, social studies or process optimization. The aim of this thesis is therefore to develop a system for the identification and labeling of persons and groups.
Entwicklung eines digitalen Zwillings für eine Tello-Drohne im MARS-Framework
Author: Leon Chun Wai Yuen
Type: Bachelor's thesis
Semester: Summer semester 2023
Description: This thesis deals with the development of a digital twin prototype of a Tello-drone in a multi-agent system. A concept is presented on how a digital twin for individual quadrocopters can be developed as part of a multi-agent system within the framework of MARS.
Ein agentenbasiertes Simulationsmodell für die Ausbreitung von Wiedehopfen in Norddeutschland
Author: David Duong
Type: Bachelor's thesis
Semester: Summer semester 2023
Description: This present work covers the development of multi-agent system, with the purpose of predicting the distribution of hoopoes in north germany. The hoopoes are agents and are using georeferenced data to accurately predict potential habitats.
Bestimmung einer Gruenen Welle bei Lichtsignalschaltungen für Alster-Fahrradfahrer durch agentenbasierte Simulation mithilfe des MARS-Frameworks
Author: Kalvin Doege
Type: Bachelor's thesis
Semester: Summer semester 2022
Description: This study addresses the question of whether and under what conditions it would be possible to achieve a green wave for cyclists in Hamburg's city center. With the help of an agent-based traffic model, experiments were conducted and evaluated against the question.
Implementierung einer ereignisbasierten Routenanpassung in MARS mithilfe von simulierten IoT-Sensordaten
Author: Daniel Osterholz
Type: Bachelor's thesis
Semester: Winter semester 2021
Description: The work addresses the question of whether it is possible to achieve a reduction in inner-city road traffic by using IoT sensor data. With the help of suitable scenarios, a statement is then made about the effectiveness of IoT sensors in parking space search traffic.
Comparing Theories of Human Behaviour by Implementing them in MARS Agents: An Interdisciplinary Approach Based on the HuB-CC Framework
Author: Nima Ahmady-Moghaddam
Type: Bachelor's thesis
Semester: Winter semester 2021
Description: This thesis is concerned with the implementation of theories of human behavior as well as an agent architecture informed by domain expertise in an agent-based model written in MARS. A conceptual modeling framework called HuB-CC was used to identify, classify, and select the theories. The modeling approach was developed in an interdisciplinary exchange with the authors of the HuB-CC framework. The model results as well as the quality of the designed architecture are analyzed and potential avenues of further inquiry are explored.