Open Access

This article introduces a model-based design, implementation, deployment, and execution methodology, with tools supporting the systematic composition of algorithms from generic and domain-specific computational building blocks that prevent code duplication and enable robots to adapt their software themselves. The envisaged algorithms are numerical solvers based on graph structures. In this article, we focus on kinematics and dynamics algorithms, but examples such as message passing on probabilistic networks and factor graphs or cascade control diagrams fall under the same pattern. The tools rely on mature standards from the Semantic Web. They first synthesize algorithms symbolically, from which they then generate efficient code. The use case is an overactuated mobile robot with two redundant arms.

Grading student answers and providing feedback are essential yet time-consuming tasks for educators. Recent advancements in Large Language Models (LLMs), including ChatGPT, Llama, and Mistral, have paved the way for automated support in this domain. This paper investigates the efficacy of instruction-following LLMs in adhering to predefined rubrics for evaluating student answers and delivering meaningful feedback. Leveraging the Mohler dataset and a custom German dataset, we evaluate various models, from commercial ones like ChatGPT to smaller open-source options like Llama, Mistral, and Command R. Additionally, we explore the impact of temperature parameters and techniques such as few-shot prompting. Surprisingly, while few-shot prompting enhances grading accuracy closer to ground truth, it introduces model inconsistency. Furthermore, some models exhibit non-deterministic behavior even at near-zero temperature settings. Our findings highlight the importance of rubrics in enhancing the interpretability of model outputs and fostering consistency in grading practices.

The workshop XAI for U aims to address the critical need for transparency in Artificial Intelligence (AI) systems that integrate into our daily lives through mobile systems, wearables, and smart environments. Despite advances in AI, many of these systems remain opaque, making it difficult for users, developers, and stakeholders to verify their reliability and correctness. This workshop addresses the pressing need for enabling Explainable AI (XAI) tools within Ubiquitous and Wearable Computing and highlights the unique challenges that come with it, such as XAI that deals with time-series and multimodal data, XAI that explains interconnected machine learning (ML) components, and XAI that provides user-centered explanations. The workshop aims to foster collaboration among researchers in related domains, share recent advancements, address open challenges, and propose future research directions to improve the applicability and development of XAI in Ubiquitous Pervasive and Wearable Computing - and with that seeks to enhance user trust, understanding, interaction, and adoption, ensuring that AI- driven solutions are not only more explainable but also more aligned with ethical standards and user expectations.

Few mobile robot developers already test their software on simulated robots in virtual environments or sceneries. However, the majority still shy away from simulation-based test campaigns because it remains challenging to specify and execute suitable testing scenarios, that is, models of the environment and the robots’ tasks. Through developer interviews, we identified that managing the enormous variability of testing scenarios is a major barrier to the application of simulation-based testing in robotics. Furthermore, traditional CAD or 3D-modelling tools such as SolidWorks, 3ds Max, or Blender are not suitable for specifying sceneries that vary significantly and serve different testing objectives. For some testing campaigns, it is required that the scenery replicates the dynamic (e.g., opening doors) and static features of real-world environments, whereas for others, simplified scenery is sufficient. Similarly, the task and mission specifications used for simulation-based testing range from simple point-to-point navigation tasks to more elaborate tasks that require advanced deliberation and decision-making. We propose the concept of composable and executable scenarios and associated tooling to support developers in specifying, reusing, and executing scenarios for the simulation-based testing of robotic systems. Our approach differs from traditional approaches in that it offers a means of creating scenarios that allow the addition of new semantics (e.g., dynamic elements such as doors or varying task specifications) to existing models without altering them. Thus, we can systematically construct richer scenarios that remain manageable. We evaluated our approach in a small simulation-based testing campaign, with scenarios defined around the navigation stack of a mobile robot. The scenarios gradually increased in complexity, composing new features into the scenery of previous scenarios. Our evaluation demonstrated how our approach can facilitate the reuse of models and revealed the presence of errors in the configuration of the publicly available navigation stack of our SUT, which had gone unnoticed despite its frequent use.

During robot-assisted therapy, a robot typically needs to be partially or fully controlled by therapists, for instance using a Wizard-of-Oz protocol; this makes therapeutic sessions tedious to conduct, as therapists cannot fully focus on the interaction with the person under therapy. In this work, we develop a learning-based behaviour model that can be used to increase the autonomy of a robot’s decision-making process. We investigate reinforcement learning as a model training technique and compare different reward functions that consider a user’s engagement and activity performance. We also analyse various strategies that aim to make the learning process more tractable, namely i) behaviour model training with a learned user model, ii) policy transfer between user groups, and iii) policy learning from expert feedback. We demonstrate that policy transfer can significantly speed up the policy learning process, although the reward function has an important effect on the actions that a robot can choose. Although the main focus of this paper is the personalisation pipeline itself, we further evaluate the learned behaviour models in a small-scale real-world feasibility study in which six users participated in a sequence learning game with an assistive robot. The results of this study seem to suggest that learning from guidance may result in the most adequate policies in terms of increasing the engagement and game performance of users, but a large-scale user study is needed to verify the validity of that observation.

This work proposes a novel approach for probabilistic end-to-end all-sky imager-based nowcasting with horizons of up to 30 min using an ImageNet pre-trained deep neural network. The method involves a two-stage approach. First, a backbone model is trained to estimate the irradiance from all-sky imager (ASI) images. The model is then extended and retrained on image and parameter sequences for forecasting. An open access data set is used for training and evaluation. We investigated the impact of simultaneously considering global horizontal (GHI), direct normal (DNI), and diffuse horizontal irradiance (DHI) on training time and forecast performance as well as the effect of adding parameters describing the irradiance variability proposed in the literature. The backbone model estimates current GHI with an RMSE and MAE of 58.06 and 29.33 W m−2, respectively. When extended for forecasting, the model achieves an overall positive skill score reaching 18.6 % compared to a smart persistence forecast. Minor modifications to the deterministic backbone and forecasting models enables the architecture to output an asymmetrical probability distribution and reduces training time while leading to similar errors for the backbone models. Investigating the impact of variability parameters shows that they reduce training time but have no significant impact on the GHI forecasting performance for both deterministic and probabilistic forecasting while simultaneously forecasting GHI, DNI, and DHI reduces the forecast performance.