This thesis addresses the problem of action recognition through the analysis of human motion and the benchmarking of its imitation by robotic systems. For our action recognition related approaches, we focus on presenting approaches that generalize well across different sensor modalities. We transform multivariate signal streams from various sensors to a common image representation. The action recognition problem on sequential multivariate signal streams can then be reduced to an image classification task for which we utilize recent advances in machine learning. We demonstrate the broad applicability of our approaches formulated as a supervised classification task for action recognition, a semi-supervised classification task for one-shot action recognition, modality fusion and temporal action segmentation. For action classification, we use an EfficientNet Convolutional Neural Network (CNN) model to classify the image representations of various data modalities. Further, we present approaches for filtering and the fusion of various modalities on a representation level. We extend the approach to be applicable for semi-supervised classification and train a metric-learning model that encodes action similarity. During training, the encoder optimizes the distances in embedding space for self-, positive- and negative-pair similarities. The resulting encoder allows estimating action similarity by calculating distances in embedding space. At training time, no action classes from the test set are used. Graph Convolutional Network (GCN) generalized the concept of CNNs to non-Euclidean data structures and showed great success for action recognition directly operating on spatio-temporal sequences like skeleton sequences. GCNs have recently shown state-of-the-art performance for skeleton-based action recognition but are currently widely neglected as the foundation for the fusion of various sensor modalities. We propose incorporating additional modalities, like inertial measurements or RGB features, into a skeleton-graph, by proposing fusion on two different dimensionality levels. On a channel dimension, modalities are fused by introducing additional node attributes. On a spatial dimension, additional nodes are incorporated into the skeleton-graph. Transformer models showed excellent performance in the analysis of sequential data. We formulate the temporal action segmentation task as an object detection task and use a detection transformer model on our proposed motion image representations. Experiments for our action recognition related approaches are executed on large-scale publicly available datasets. Our approaches for action recognition for various modalities, action recognition by fusion of various modalities, and one-shot action recognition demonstrate state-of-the-art results on some datasets. Finally, we present a hybrid imitation learning benchmark. The benchmark consists of a dataset, metrics, and a simulator integration. The dataset contains RGB-D image sequences of humans performing movements and executing manipulation tasks, as well as the corresponding ground truth. The RGB-D camera is calibrated against a motion-capturing system, and the resulting sequences serve as input for imitation learning approaches. The resulting policy is then executed in the simulated environment on different robots. We propose two metrics to assess the quality of the imitation. The trajectory metric gives insights into how close the execution was to the demonstration. The effect metric describes how close the final state was reached according to the demonstration. The Simitate benchmark can improve the comparability of imitation learning approaches.

Teilgrundordnung für das Qualitätssicherungssystem nach § 5 HochSchG an der Universität Koblenz Ordnung zur Qualitätssicherung und -entwicklung in Studium und Lehre an der Universität Koblenz Sechzehnte Ordnung zur Änderung der Prüfungsordnung für die Prüfung im lehramtsbezogenen Bachelorstudiengang Berufsbildende Schulen an der Universität Koblenz-Landau, der Hochschule Koblenz und der Philosophisch-Theologischen Hochschule Vallendar Siebzehnte Ordnung zur Änderung der Prüfungsordnung für die Prüfung im lehramtsbezogenen Bachelorstudiengang Berufsbildende Schulen an der Universität Koblenz-Landau, der Hochschule Koblenz und der Philosophisch-Theologischen Hochschule Vallendar Fünfzehnte Ordnung zur Änderung der Ordnung für die Prüfung im Masterstudiengang Lehramt an berufsbildenden Schulen an der Universität Koblenz-Landau, der Hochschule Koblenz und der Philosophisch-Theologischen Hochschule Vallendar Sechszehnte Ordnung zur Änderung der Ordnung für die Prüfung im Masterstudiengang Lehramt an berufsbildenden Schulen an der Universität Koblenz-Landau, der Hochschule Koblenz und der Philosophisch-Theologischen Hochschule Vallendar Sechsundzwanzigste Ordnung zur Änderung der Ordnung für die Prüfung im lehramtsbezogenen Zertifikatsstudiengang (Erweiterungsprüfung) an der Universität Koblenz-Landau, Campus Koblenz und der Hochschule Koblenz Siebenundzwanzigste Ordnung zur Änderung der Ordnung für die Prüfung im lehramtsbezogenen Zertifikatsstudiengang (Erweiterungsprüfung) an der Universität Koblenz-Landau, Campus Koblenz und der Hochschule Koblenz Richtlinie der Universität Koblenz zur Einrichtung, Ausschreibung und Besetzung von Professuren und Juniorprofessuren (Berufungsrichtlinie) Satzung und Qualitätssicherungskonzept der Universität Koblenz zur Einrichtung, Ausschreibung und Besetzung von Juniorprofessuren (W1) mit und ohne Tenure-Track sowie zur Besetzung von Professuren unter Verzicht auf eine Ausschreibung (Tenure-Satzung) Benutzungsordnung der Universitätsbibliothek Bibliotheksordnung der Universität Koblenz

Redaktionelle Korrektur betreffend die Zweite Ordnung zur Änderung der Gemeinsamen Prüfungsordnung für die Bachelor- und Masterstudiengänge des Fachbereichs Informatik an der Universität Koblenz-Landau vom 03. August 2022 Redaktionelle Korrektur betreffend die Dreiundzwanzigste Ordnung zur Änderung der Prüfungsordnung für die Prüfung im Zwei-Fach-Bachelorstudiengang an der Universität Koblenz-Landau, Campus Koblenz vom 19. September 2022 Einschreibeordnung der Rheinland-Pfälzischen Technischen Universität Kaiserslautern-Landau Satzung der Rheinland-Pfälzisch Technische Universität Kaiserslautern-Landau über das Auswahlverfahren in zulassungsbeschränkten Studiengängen (Auswahlsatzung)

Ponds in agricultural landscapes are often used by amphibians as breeding habitat. However, the characteristics of agricultural ponds and especially the surrounding area are usually said to be suboptimal for many amphibian species. Using suboptimal habitats might allow a species’ survival and reproduction, but can have negative consequences at the individual and population level. In the present study, we investigated Palmate Newt (Lissotriton helveticus) populations from an intensive wine-growing region in southern Germany and compared them with populations located in a nearby forested area in terms of biometric traits, age and genetic structure. By analyzing over 900 adult newts from 11 ponds, we could show that newts reproducing in forest ponds were larger than newts reproducing in agricultural ponds. We did not find differences in the newt age and growth rate between habitat types. Therefore, differences in the body size of newts might already existed in larvae and/or juveniles, what might be related to a lower habitat quality for larvae and/or juveniles in the agricultural landscape. Body mass, body condition and sexual dimorphic traits (length of the caudal filament and max. height of the tail) correlated with body size, but no additional effect of the habitat type was found. The analysis of microsatellites revealed a higher genetic diversity in forest ponds. However, no clear sign of inbreeding was observed in any agricultural population, suggesting some degree of gene flow between them. We conclude, that agricultural ponds can be suitable habitats for the Palmate Newt and that conservation effort should aim to preserve them. The observed effects on body size indicate the need to increase the quality of the aquatic and terrestrial habitat for early life stages of this newt species in agricultural landscapes.

This thesis was motivated by the need to advance the knowledge on the variability and dynamics of energy fluxes in lakes and reservoirs, as well as about the physical processes that regulate the fluxes at both the air and water side of the air-water interface. In the first part, I re-examine how mechanical energy, resulting from its major source – the vertical wind energy flux - distributes into the various types of water motions, including turbulent flows and surface and internal waves. Although only a small fraction of the wind energy flux from the atmosphere is transferred to the water, it is crucial for physical, biogeochemical and ecological processes in lentic ecosystems. Based on extensive air- and water-side measurements collected at two small water bodies (< 10 km2), we estimated the energy fluxes and energy content in surface and in internal waves. Overall, the estimated energy fluxes and energy content agree well with results reported for larger water bodies, suggesting that the energetics driving the water motions in enclosed basins is similar, independently of the basin size. Our findings highlight the importance of the surface waves that receive the largest fraction of the wind energy flux, which strongly nonlinearly increases for wind speeds exceeding 3 m s-1. We found that the existing parameterization of the wave height as a function of wind speed and fetch length did not reproduce the measured wave amplitude in lakes. On average, the highest energy content was observed in basin-scale internal waves, together with high-frequency internal waves exhibiting seasonal variability and varies among the aquatic systems. During our analysis, we discovered the diurnal variability of the energy dissipation rates in the studied lake, suggesting biogenic turbulence generation, which appears to be a widespread phenomenon in lakes and reservoirs. In the second part of the thesis, I addressed current knowledge gaps related to the bulk transfer coefficients (also known as the drag coefficient, the Stanton number and the Dalton number), which are of a particular importance for the bulk estimation of the surface turbulent fluxes of momentum, sensible and latent heat in the atmospheric boundary layer. Their inaccurate representation may lead to significant errors in flux estimates, affecting, for example, the weather and climate predictions or estimations of the near-surface current velocities in lake hydrodynamic models. Although the bulk transfer coefficients have been extensively studied over the past several decades (mainly in marine and large-lake environments), there has been no systematic analysis of measurements obtained at lakes of different size. A significant increase of the transfer coefficients at low wind speeds (< 3 m s-1) has been observed in several studies, but, to date, it has remained unexplained. We evaluated the bulk transfer coefficients using flux measurements from 31 lakes and reservoirs. The estimates were generally within the range reported in previous studies for large lakes and oceans. All transfer coefficients increased substantially at low wind speeds, which was found to be associated with the presence of gusts and capillary waves (except the Dalton number). We found that the Stanton number is systematically higher than the Dalton number. This challenges the assumption made in the Bowen-ratio method, which is widely used for estimating evaporation rates from micrometeorological measurements. We found that the variability of the transfer coefficients among the lakes could be associated with lake surface area. In flux parameterizations at lake surfaces, it is recommended to consider variations in the drag coefficient and the Stanton number due to wind gustiness and capillary wave roughness while the Dalton number could be considered as constant at all wind speeds. In the third part of the thesis, I address the key drivers of the near-surface turbulence that control the gas exchange in a large regulated river. As all inland waters, rivers are an important natural source of greenhouse gases. The effects of the widespread alteration and regulation of river flow for human demands on gas exchange is largely unknown. In particular, the near-surface turbulence in regulated rivers has been rarely measured and its drivers have not been identified. While in lakes and reservoirs, near-surface turbulence is mainly related to atmospheric forcing, in shallow rivers and streams it is generated by bottom friction of the gravity-forced flow. The studied large regulated river represents a transition between these two cases. Atmospheric forcing and gravity were the dominant drivers of the near-surface turbulence for a similar fraction of the measurement period. Based on validated scalings, we derived a simple model for estimating the relative contributions of wind and bottom friction to near-surface turbulence in lotic ecosystems with different flow depths. Large diel variability in the near-surface energy dissipation rates due to flow regulation leads to the same variability in gas exchange. This suggests that estimates of gas fluxes from rivers are biased by measurements performed predominantly during daytime. In addition, the novelty in all the analyses described above is the use of the turbulent surface fluxes measured directly by the eddy-covariance technique – at the moment of writing, the most advanced method. Overall, this thesis is of a potential interest for a broad range of scientific disciplines, including limnology, micrometeorology and open channel hydraulics.