Recent Submissions
Optimising light electric vehicle capabilities through a new battery main switch topology with emphasis on battery heating
(2025) Oehl, Joachim; Birke, Kai Peter (Prof. Dr.-Ing.)
Tunable muscular damping to increase morphological computation in legged locomotion
(Stuttgart : Institut für Modellierung und Simulation Biomechanischer Systeme, Computational Biophysics and Biorobotics, 2025) Araz, Matthew; Schmitt, Syn (Prof. Dr.)
Legged locomotion is a complex motor control task that animals have to perform efficiently and adaptively to survive in nature. It has been shown that animals often adopt compliant limbs to achieve energy-efficient locomotion, where they conserve and recycle energy. Meanwhile, they are also highly robust over uneven environments where unexpected perturbations occur, even though the nervous system fails to generate instant feedback for stability. Although robotics researchers often state energy efficiency and robustness as contradicting characteristics of legged locomotion, since efficiency requires energy conservation and robustness requires energy dissipation, animals are remarkably capable of achieving these two locomotion characteristics simultaneously. Early studies suggested that when animals encounter an unexpected perturbation, muscle intrinsic mechanics generate an instant mechanical feedback, also known as preflexes, within the time frame of neural transmission delays, which contributes to the perturbation rejection and robustness. Muscle preflexes are the force output of viscoelastic mechanical properties of the muscle, which arise from the changes in length and contraction velocity of the muscle. It has been shown that muscles tend to dissipate energy when they are elongated under external loads. Several studies suggested that this morphological capacity of muscles might be due to their viscoelastic structure, which might play a critical role in generating energetically efficient and robust locomotion simultaneously and reduce the reflex-mediated control effort. Yet, how animals leverage their intrinsic muscle mechanics to generate both energy-efficient and robust locomotion is still not well understood. This dissertation explores how animals benefit from their intrinsic muscle mechanics to achieve energy-efficient and robust locomotion against step perturbations. The research focused on the phenomenological velocity-dependent force behavior of muscles, which allows the energy dissipation, in the early stance, the factors influencing muscles’ viscous-like damping capacity, and the potential benefits of this physiological characteristic to locomotion. This dissertation hypothesized that muscles’ intrinsic mechanics can generate an instant force to the stretch velocities induced by perturbations and dissipate some part of the perturbation energy before the reflex-mediated control is activated. Besides, this viscous-like damping capacity of the muscles is tunable by the muscle activity level, which turns muscles into tunable dampers that operate as embedded low-level controllers within the animal’s morphology. This dissertation investigates the main hypothesis through three contributions (one peer-reviewed journal article, one submitted manuscript to a peer-reviewed journal, and one manuscript in preparation to be submitted to a peer-reviewed journal) that I have worked on throughout my doctoral project.
The muscles’ viscous-like tunable damping behavior and its potential benefits to locomotion were explored based on two different methodologies: in vitro experiments and neuromusculoskeletal simulation analysis. The first contribution of this dissertation suggested a novel approach to test biological muscle fibers in vitro under physiologically realistic boundary conditions, which are obtained from neuromusculoskeletal simulations. This approach allowed us to test muscular behavior at real-life like impact conditions that occur due to unexpected perturbations. The main finding of this contribution showed that biological muscles generate an instant viscoelastic reaction to the impacts that occur due to the perturbations. While the initial phase of this reaction is elastic, the following phase is altered depending on the stretch velocities induced by step perturbations. Besides, we observed that changing the muscle activity level tunes the muscular impedance. This shows that muscles operate like tunable muscular damping, and their damping capacity can be altered via feedforward neural modulation. Yet, under constant activity, the amount of stretch is found to be the driving factor of main energy dissipation, not the muscular damping.
In addition to these results, the first contribution also provided a comparison between the biological muscle’s and the Hill-type muscle model’s preflex behavior for model validation. This comparison showed that the Hill-type muscle model can predict preflex work with an acceptable accuracy, even though it fails to predict the force behavior of the biological muscle fibers. This shows that the Hill-type muscle model can still be a useful tool to investigate muscle preflex work under conditions that are not possible to test in vitro or in vivo. Thus, the other two contributions of this dissertation used neuromusculoskeletal hopping simulations, that actuated by a Hill-type muscle model, to investigate the factors influencing the muscle preflexes and the contribution of muscle preflexes to generating energy-efficient and robust locomotion simultaneously.
While the first contribution showed that the muscle’s tunable damping characteristic is not the main contributor of energy dissipation under constant activity, the second contribution showed that neural modulation by feedforward stimulation is necessary to activate the force-velocity relationship and position it as the primary component of energy dissipation. The effectiveness of the enhanced muscular damping response is sensitive to the complex interplay between tendon stiffness and prestimulation level. Furthermore, interaction with the environment makes this inner mechanical tuning more robust and allows effective use of enhanced damping-like capacity.
Building on the first two contributions that focused on understanding the preflex mechanics and the factors influencing its effectiveness, the third contribution of this dissertation investigated the trade-off between energy efficiency and robustness at muscle-actuated locomotion. Analysis of this contribution showed that muscle intrinsic mechanics play a crucial role in obtaining energy-efficient and robust locomotion simultaneously. Preplanned feed-forward stimulation optimized to energetically efficient use of the muscle-tendon unit promotes near-optimal robustness against step-down perturbations. To generate an energetically efficient gait, animals benefit from the elastic structure of the tendons by operating muscles around near isometric and recycling the energy via tendons. This requires minimizing muscle length during steady locomotion, which can be obtained by actively stiffening the muscle against impact forces. Thus, muscle fibers generate high adaptability to the perturbations with effective use of tunable muscular damping. This outcome shows that morphological characteristics of the muscle-tendon unit resolve the trade-off between energy efficiency and robustness in biological locomotion.
In summary, this dissertation enhances the current understanding of muscle preflex behavior during perturbation rejection and the potential contribution of this phenomenological characteristic to generate versatile locomotion. The findings of the contributions prove the main hypothesis of this dissertation, that the intrinsic muscle mechanics operate like a tunable muscular damper that provides an instant dissipative reaction to the sudden changes in system dynamics and reduces the control effort of the delayed reflex-mediated feedback control. This indicates that muscles function as embedded low-level controllers, simplifying movement control for the high-level controller and promoting energy-efficient and robust locomotion. Understanding the mechanics of this morphological capacity is crucial for gaining insight into the mechanisms underlying the motor control of animal movement. Furthermore, the reverse engineering of muscular intrinsic mechanics and the production of muscle-tendon actuators may enable us to generate highly versatile and efficient robots that can operate like animals in nature.
Overview on web authentication and authorization protocol security evaluations
(2025) Schäberle, Victoria
Web authentication and authorization protocols such as OAuth 2.0 and OpenID Connect are widely used and have become established standards in the single sign-on (SSO) ecosystem. They enable users to grant third-party applications access to their resources and to use diverse online services conveniently and securely, while reducing the risks associated with password-based authentication. However, incorrect implementations have repeatedly led to severe vulnerabilities, ranging from token theft to account compromise. These risks highlight the importance of understanding the security properties and weaknesses of these protocols.
This thesis investigates the security of OAuth 2.0 and OpenID Connect by providing a structured overview of existing evaluations, considering both foundational and widely cited studies as well as more recent contributions. Both scientific and non-scientific sources are considered, including formal analyses, empirical studies, and protocol specifications. Typical attacks and vulnerabilities are summarized and categorized into four fields, which include formal, empirical, protection-centered, and investigations of specific attacks and mitigations. The review highlights approaches that develop tools for empirical evaluation, focus on mobile applications, or analyze relying parties (RPs) and identity providers (IdPs). The methodology of the empirical studies examined range from active and passive attack strategies to white-box and black-box analyses.
By synthesizing these findings, this thesis reveals a persistent gap between the formal security of the protocols and recurring weaknesses in real-world implementations. This gap underscores the need for continued research, clearer developer guidance, and systematic auditing.
Homotopie des rotierenden Stabes
(2020) Mair, Jonas
Optimal leg-stiffness in a 2D monoped
(2024) Tschemernjak, Dominik
Tracing optimal gaits of the compass-gait biped
(2021) Katamish, Bassel
A benchmark study on problems related to CO2 storage in geologic formations : summary and discussion of the results
(2009) Class, Holger; Ebigbo, Anozie; Helmig, Rainer; Dahle, Helge; Nordbotten, Jan M.; Celia, Michael A.; Audigane, Pascal; Darcis, Melanie; Ennis-King, Jonathan; Fan, Yaqing; Flemisch, Bernd; Gasda, Sarah; Jin, Min; Krug, Stefanie; Labregere, Diane; Naderi, Ali; Pawar, Rajesh J.; Sbai, Adil; Thomas, Sunil G.; Trenty, Laurent; Wei, Lingli
This paper summarises the results of a benchmark study that compares a number of mathematical and numerical models applied to specific problems in the context of carbon dioxide (CO2) storage in geologic formations. The processes modelled comprise advective multiphase flow, compositional effects due to dissolution of CO2 into the ambient brine, and non-isothermal effects due to temperature gradients and the Joule-Thompson effect. The problems deal with leakage through a leaky well, methane recovery enhanced by CO2 injection, and a reservoir-scale injection scenario into a heterogeneous formation.
We give a description of the benchmark problems, then briefly introduce the participating codes, and finally present and discuss the results of the benchmark study.
Vorberechnung und Darstellung von POI-Labelling-Hierarchien
(2025) Zelia, Mykhailo
Diese Arbeit befasst sich mit der automatischen Platzierung von Point-of-Interest (POI)-Labels in digitalen Karten. Dazu wurde das Wachstumsmodell für Kreise von Funke et al. auf achsenparallele Rechtecke übertragen, um die Form von Textlabels realitätsnäher zu berücksichtigen. Die zentrale Herausforderung besteht darin, möglichst viele Labels darzustellen, Überlappungen zu vermeiden, Konsistenz über verschiedene Zoomstufen hinweg sicherzustellen und gleichzeitig Verfahren zu entwickeln, die auch auf sehr großen Datensätzen wie OpenStreetMap (OSM) effizient einsetzbar sind.
Theoretisch konnten obere Laufzeitschranken für zwei Varianten von dem Algorithmus (Chebyshev-Nearest-Neighbor- und Euclidean-Nearest-Neighbor-basiert) unter Einsatz kombinierter Datenstrukturen hergeleitet werden. Für den praktischen Einsatz ergibt sich dabei eine effektive Laufzeitvon 𝑂(𝑛 log^4 𝑛), wobei der Speicherbedarf 𝑂(𝑛 log 𝑛) für die CNN-Variante und 𝑂(𝑛) für die ENN-Variante beträgt. Beide Ansätze wurden implementiert und in Verbindung mit einem Quadtree ausgewertet.
Die Experimente zeigten, dass beide Verfahren identische Ergebnisse liefern, die ENN-Variante jedoch schnellere Berechnungszeiten erreicht. Für den Planetendatensatz mit 12,5 Millionen Labels konnten die Kollisionszeiten in weniger als 25 Minuten berechnet werden. Ein anschließender Webserver-Prototyp zur Abfrage der vorberechneten Daten zeigte, dass Suchanfragen in wenigen Millisekunden beantwortet werden können.
Damit leistet die Arbeit sowohl einen theoretischen als auch einen praktischen Beitrag zur effizienten und skalierbaren Vorberechnung von Labelhierarchien in interaktiven Kartendiensten.
Gender inclusive occupational titles in Wikipedia and large language models
(2025) Caulfield, Emma
This thesis explores the use of gender-inclusive occupational titles across different languages in Wikipedia and by prompt-based language models. Wikipedia, being a publicly editable and widely accessed online encyclopedia, serves as a reflection of everyday language use. Additionally, the thesis investigates the correlation between occupational titles in Wikipedia and those generated by language models. The analysis encompasses four languages: English, German, Spanish, and Italian, selected for their representation on Wikipedia and their varied approaches to inclusive language. The findings indicate that both Wikipedia authors and language models tend to use inclusive forms of occupational titles, with language models exhibiting a higher tendency to use inclusive forms to refer to women. This trend suggests that language models may be mirroring broader linguistic shifts towards gender-inclusive language, and potentially promoting more inclusive language use. As influential platforms like Wikipedia shape both public perception and language model training, the presence of gender-inclusive language in these sources may contribute to the normalization of such practices in wider communication.
Comparison of distributional and visual nearest neighbors
(2025) Naber, Sven
This thesis investigates how semantic concepts are represented across textual and visual embedding spaces, focusing on the abstract-concrete continuum. Using 5,448 English nouns and their embeddings from both distributional language models (e.g., Word2Vec, GloVe) and vision models (e.g., ViT, DINOv2, CLIP), it compares neighborhood structure via a normalized alignment score (NAS). Results show that alignment is primarily driven by input modality rather than model architecture, with strong local overlap for concrete concepts and more diffuse agreement for abstract ones. Mean aggregation of image embeddings improves visual consistency but cannot fully bridge modality-specific limitations. The findings provide a starting point for further exploration of semantic spaces.