Julius Schöning

In smart city data engineering, accurate traffic forecasting is crucial for effective urban planning and management, as well as for developing Urban Digital Twins (UDTs). Extending prior work [1] on multistep, multi-scale traffic flow forecasting, this paper explores integrating weather data into traffic forecasting models using graph neural networks (GNNs) to assess its impact. The study focuses on selecting features and time ranges for multivariate multistep traffic forecasting. It experiments with GNN setups and compares single-node per Traffic Eye Universal (TEU) configurations with combined nodes. The impact of incorporating weather features on model accuracy and training efficiency is evaluated. In the present setting, with motorized traffic in the city of Osnabrück and weather data available only at the city level, the key findings are that weather data did not improve forecasting accuracy and increased training time. Potential benefits for bicycles and pedestrians are noted as a hypothesis, as data limitations preclude further exploration. The study underscores the efficacy of correlation matrices and mutual information for feature selection, highlighting the importance of capturing non-linear relationships. It suggests that, in this configuration, the exclusive use of TEU data yields the best performance, with the node structure having a negligible impact on results. These findings imply that city-level weather data may not be a predictive value for all traffic types and motivate future research on sensor-level weather and road-surface data, as well as richer feature selection. This study contributes to the understanding of smart city data systems, underscoring the importance of efficient feature selection and the challenges of integrating diverse data sources.

Assessing child endangerment in the home environment can be a substantial challenge. It is particularly challenging for social workers with little experience in assessment of risks for child endangerment. The study examines the current knowledge and application of VR-based technologies in training professionals to assess child endangerment, including how these interventions are developed and evaluated using specific tools. It also explores the effects of VR on skills development and participant perception compared to traditional methods and the extent to which they are employed in decision-making for child welfare. A scoping review based on PRISMA-ScR and the Arksey and O’Malley framework was conducted, and seven scientific databases were systematically searched from 2014 to 2024. Of 307 identified studies, ten remained after screening. VR is already used for the training of professionals but is hardly incorporated in the field of social work. VR training is more realistic and immersive than conventional training methods. However, its effectiveness also depends on variables such as technologies used and the incorporation of feedback. Our review indicates that VR serves as an effective, immersive instructional tool for developing essential skills in child welfare risk assessments, offering advantages over traditional methods in realism and user engagement. However, challenges remain in standardizing VR technologies and evaluating their real-world decision-making impacts. While VR-based methods seem to be promising for the assessment training of child endangerment, scientific research is still sparse and does not provide conclusive results for all investigated outcomes.

Accurate traffic flow forecasting is critical for urban mobility management. Nonetheless, existing methods often face challenges in multi-step prediction scenarios over extended time horizons. We present a novel multi-scale traffic flow forecasting framework using graph neural networks, which are capable of predicting traffic patterns of up to three days ahead at varying temporal resolutions. We extend Spatiotemporal Self-Supervised Learning with methods to integrate multi-scale data processing, combining high-resolution scale with low-resolution scales to balance computational efficiency and accuracy. We leverage our own dataset, comprising static network topology and dynamic sensor data, and demonstrate that synthetic adjacency matrices can replace handcrafted ones without significantly sacrificing performance, streamlining the preprocessing. Key innovations include a scale-space representation for temporal data and a network architecture adapted to non-uniform time series. Further, our results show that preprocessing significantly enhances forecasting accuracy while multi-step prediction achieves comparable accuracy results to single-step benchmarks. Our framework’s ability to handle extended forecasting horizons with reduced computational overhead highlights its practicality for real-world application, such as adaptive traffic control and logistics planning. This work advances spatial-temporal forecasting by addressing hardware constraints through multi-scale representations and lays the groundwork for future extensions incorporating multimodal data and expert knowledge.

Synthetic data has emerged as a cost-effective alternative to real data for training artificial neural networks (ANN). However, the disparity between synthetic and real data results in a domain gap. That gap leads to poor performance and generalization of the trained ANN when applied to real-world scenarios. Several strategies have been developed to bridge this gap, which combine synthetic and real data, known as mixed training using hybrid datasets. While these strategies have been shown to mitigate the domain gap, a systematic evaluation of their generalizability and robustness across various tasks and architectures remains underexplored. To address this challenge, our study comprehensively analyzes two widely used mixing strategies on three prevalent architectures and three distinct hybrid datasets. From these datasets, we sample subsets with varying proportions of synthetic to real data to investigate the impact of synthetic and real components. The findings of this paper provide valuable insights into optimizing the use of synthetic data in the training process of any ANN, contributing to enhancing robustness and efficacy.

The agricultural industry is undergoing a significant transformation with the increasing adoption of autonomous technologies. Addressing complex challenges related to safety and security, components and validation procedures, and liability distribution is essential to facilitate the adoption of autonomous technologies. This paper explores the collaborative groups and initiatives undertaken to address these challenges. These groups investigate inter alia three focal topics: 1) describe the functional architecture of the operational range, 2) define the work context, i.e., the realistic scenarios that emerge in various agricultural applications, and 3) the static and dynamic detection cases that need to be detected by sensor sets. Linked by the Agricultural Operational Design Domain (Agri-ODD), use case descriptions, risk analysis, and questions of liability can be handled. By providing an overview of these collaborative initiatives, this paper aims to highlight the joint development of autonomous agricultural systems that enhance the overall efficiency of farming operations.