To Örebro University

The integration of Artificial Intelligence (AI) in Cyber-Physical Production Systems (CPPS) is rapidly transforming modern manufacturing. AI is becoming instrumental in improving operational efficiency, decision-making, and robotic service orchestration. However, the opaque nature of AI decision-making introduces significant risks in environments where AI-driven decisions can have physical consequences. This study proposes a novel Model-Driven Service-Oriented Adaptive Data Basin framework (AI.Lock), which applies strict control over the data accessible to AI engines and limits their actions within predefined, model-driven boundaries. By adopting a model-based architecture, this framework enforces data restriction to a need-to-know basis for AI engines, ensuring that only relevant data subsets are accessible based on the service or task at hand. Additionally, it encapsulates AI behaviours within the scope of predefined models, mitigating safety risks and ensuring that actions remain predictable, traceable, and compliant with factory operations. The framework is particularly suited for multi-party, and multi-vendor environments, where model-driven approaches provide verifiable constraints. Furthermore, the proposed framework leverages Open Platform Communications Unified Architecture's (OPC-UA) communication and information modelling capabilities, integrating with digital twins to ensure seamless adaptation in dynamic CPPS environments. This study explores the benefits of using a model-driven approach to enhance safety, reduce risk, and provide a foundation for secure AI integration in modern manufacturing.

Monitoring of gas emissions is critical for ensuring safety, efficiency, and compliance in industrial plants. Approaches based on stationary sensor networks leave areas of the plant unmonitored and lack adaptability to changing conditions. Combining stationary with mobile sensors on quadruped robots offers a robust approach with improved spatial resolution, adaptability and responsiveness. To streamline the integration of such systems, a unified and open software framework is required. We propose to abstract the low-level implementations, e.g., the robot control layers, using the REST API to facilitate the implementation of high-level tasks such as data visualization and AI-based analysis, thus enabling scalable monitoring solutions. An indoor gas release experiment was conducted, which showed the benefits of the quadruped robot-based and the combined sensing approach.

We propose a flexible hierarchical navigation stack for a mobile robot in complex dynamic environments. Addressing the growing need for reliable navigation in real-world scenarios, where dynamic agents and environmental uncertainties pose significant challenges, our solution decomposes this complexity into task planning, navigation, control, and safe velocity components. In contrast to the prior art, our system at every level incorporates diverse contextual information about the environment, anticipates navigation risks and proactively avoids collisions with dynamic agents.

Gas source localization in complex environments is critical for applications such as environmental monitoring, industrial safety, and disaster response. Traditional methods often struggle with the challenges posed by a lack of environmental topography integration, especially when interactions between wind and obstacles distort gas dispersion patterns. In this paper, we propose a deep learning-based approach, which leverages spatial context and environmental mapping to enhance gas source localization. By integrating Simultaneous Localization and Mapping (SLAM) with a U-Net-based model, our method predicts the likelihood of gas source locations by analyzing gas sensor data, wind flow, and topography of the environment represented by a 2D occupancy map. We demonstrate the efficacy of our approach using a wheeled robot equipped with a photoionization detector, a LIDAR, and an anemometer, in various scenarios with dynamic wind fields and multiple obstacles. The results show that our approach can robustly locate gas sources, even in challenging environments with fluctuating wind directions, outperforming conventional methods by utilizing topography contextual information. This study underscores the importance of topographical context in gas source localization and offers a flexible and robust solution for real-world applications. Data and code are publicly available.

Eye tracking (ET) has shown great potential in mathematics education research (Schindler et al., 2020). Since high-quality ET (hqET) systems are expensive, cost-efficient alternatives are needed for practical applications, e.g., in schools. Webcam-based ET (wcET) provides an affordable alternative, yet comes with a lower data quality. Our research aims to make wcET usable for practical applications in mathematics education. To achieve this, we intend to enhance wcET data through, e.g., automatic offset calibration (Asgahri et al., 2023) and semantic remapping (using knowledge about the presented tasks), so that relevant information can be extracted from the students’ eye movement patterns as with hqET data. Here, we consider number line (NL) tasks and ask: To which degree can wcET data support math education in comparison to hqET data and how can we increase the degree?

Using our KI-ALF system (https://ki-alf.de), we collected gaze data from 136 fifth-grade students working on NL tasks using a Logitech BRIO webcam (wcET) and a Tobii Pro X3-120 eye tracker (hqET). We pre-processed the wcET data using AI for (1) correcting systematic wcET data inaccuracies using task information (automatic offset calibration); (2) grouping gaze points into areas of interest, e.g., specific numbers or landmarks, to improve data interpretation and reduce noise (semantic remapping). We evaluated the pre-processing pipeline using clustering algorithm to analyse strategy patterns in terms of cluster consistency compared the clusters to expert-labelled strategies based on hqET data. With our domain-specific post-processing, the consistency of strategy identification reached a level close to that of hqET data. Our findings show that wcET affords scalable and cost-effective solutions to identify students’ NL strategies. Based on our development, KI-ALF wcET systems are currently in use in a comprehensive school in Germany to support students’ mathematics learning utilizing eye-tracking-based identification of their strategies.

To achieve natural and intuitive interaction with people, HRI frameworks combine a wide array of methods for human perception, intention communication, human-aware navigation and collaborative action. In practice, when encountering unpredictable behavior of people or unexpected states of the environment, these frameworks may lack the ability to dynamically recognize such states, adapt and recover to resume the interaction. Large Language Models (LLMs), owing to their advanced reasoning capabilities and context retention, present a promising solution for enhancing robot adaptability. This potential, however, may not directly translate to improved interaction metrics. This paper considers a representative interaction with an industrial robot involving approach, instruction, and object manipulation, implemented in two conditions: (1) fully scripted and (2) including LLM-enhanced responses. We use gaze tracking and questionnaires to measure the participants' task efficiency, engagement, and robot perception. The results indicate higher SUbjective ratings for the LLM condition, but objective metrics show that the scripted condition performs comparably, particularly in efficiency and focus during simple tasks. We also note that the scripted condition may have an edge over LLM-enhanced responses in terms of response latency and energy consumption, especially for trivial and repetitive interactions.

LiDAR is used in autonomous driving for navigation, obstacle avoidance, and environment mapping. However, adverse weather conditions introduce noise into sensor data, potentially degrading the performance of perception algorithms and compromising the safety and reliability of autonomous driving systems. Hence, in this paper, we investigate the limitations of LiDAR perception algorithms in adverse weather conditions, explore ways to mitigate the effects of noise, and propose future research directions to achieve all-weather autonomy with LiDAR sensors. Using real-world datasets and synthetically generated dense fog, we characterize the noise in adverse weather such as snow, rain, and fog; their effect on sensor data; and how to effectively mitigate the noise for tasks like object detection, localization, and SLAM. Specifically, we investigate point cloud filtering methods and compare them based on their ability to denoise point clouds, focusing on processing time, accuracy, and limitations. Additionally, we evaluate the impact of adverse weather on state-of-the-art 3D object detection, localization, and SLAM methods, as well as the effect of point cloud filtering on the algorithms' performance. We find that point cloud filtering methods are partially successful at removing noise due to adverse weather, but must be fine-tuned for the specific LiDAR, application scenario, and type of adverse weather. 3D object detection was negatively affected by adverse weather, but performance improved with dynamic filtering algorithms. We found that heavy snowfall does not affect localization when using a map constructed in clear weather, but it fails in dense fog due to a low number of feature points. SLAM also failed in thick fog outdoors, but it performed well in heavy snowfall. Filtering algorithms have varied effects on SLAM performance depending on the type of scan-matching algorithm.

Eye tracking is gaining significance in mathematics education research at a tremendous speed. For the discipline to grow, it is essential to monitor, structure, and synthesize the research in this rapidly evolving field, which calls for a systematic literature review. However, a comprehensive and systematic review does not exist for the research for the past five years. This is a profound gap considering the dynamics of the field, which is fueled by technological advancements in hard- and software and the increasing usability and availability of eye-tracking systems. The aim of this paper is to provide a comprehensive and systematic literature review on eye-tracking research in mathematics and statistics education published in the past five years. Using a systematic database search, we identified and reviewed 116 eye-tracking studies published between 2019 and the first quarter of 2024. We found that the studies addressed a wide range of topics in all relevant curriculum content areas as well as a multitude of phenomena, including teacher-student interaction and digital learning. Interestingly, the studies increasingly involved school students, partially in authentic classroom settings. We also found that the majority of the papers referred to a theoretical framework or made assumptions about the (domain-specific) interpretation of eye movements explicit. As a further important trend, probably still in its infancy, we observed the use of AI techniques for data analysis purposes, which allows for qualitative insights despite bigger numbers of participants. Our paper provides an overview and detailed insights into trends, of which many have not been visible in earlier review studies.

Maps of dynamics are effective representations of motion patterns learned from prior observations, with recent research demonstrating their ability to enhance various downstream tasks such as human-aware robot navigation, long-term human motion prediction, and robot localization. Current advancements have primarily concentrated on methods for learning maps of human flow in environments where the flow is static, i.e., not assumed to change over time. In this paper we propose an online update method of the CLiFF-map (an advanced map of dynamics type that models motion patterns as velocity and orientation mixtures) to actively detect and adapt to human flow changes. As new observations are collected, our goal is to update a CLiFF-map to effectively and accurately integrate them, while retaining relevant historic motion patterns. The proposed online update method maintains a probabilistic representation in each observed location, updating parameters by continuously tracking sufficient statistics. In experiments using both synthetic and real-world datasets, we show that our method is able to maintain accurate representations of human motion dynamics, contributing to high performance flow-compliant planning downstream tasks, while being orders of magnitude faster than the comparable baselines. 

In this work, the Gas Source State Estimation (GSSE) problem in non-trivial geometries is approached by combining a non-stationary gas dispersion Partial Differential Equation (PDE) with in-situ gas measurements under the assumption of known flow. The GSSE problem is formulated as an optimization problem with PDE constraints and solved efficiently using the adjoint method. The approach is simulatively validated on a 2D problem with laminar flow around a circular obstacle and data from gas dispersion simulations. Thereby, a single Gaussian gas source could be localized accurately for most trials with randomly placed sensor.