Högskolan i Skövde

Discrete manufacturing is subject to challenges posed by changeovers due to diminishing batch sizes and the need for customisation. Concurrently retiring personnel deepens knowledge gaps. Despite the evident potential of AI demonstrated in various studies, the question of scalable implementation for changeovers remains largely unexplored. The present study examines the AI readiness for changeover optimisation at a global bearing manufacturer using pull-based production channel systems. The key challenges identified in this study include complex many-to-many relationships between operations and channels, products re- entering flows, and subcontracting arrangements that affect changeover efficiency. The investigation is guided by two research questions: (1) What AI capabilities provide the highest impact on changeover performance in discrete manufacturing? (2) What organisational readiness factors are necessary for successful AI solution lifecycle management? Using a literature review and case study methodology to examine model channels and changeover procedures, the study reveals significant discrepancies between AI's theoretical potential and practical realities. This work establishes a link between theoretical AI capabilities and practical implementation challenges, thus providing evidence-based guidance for firms evaluating AI opportunities. The key findings highlight that the success of the AI lifecycle depends on organisational readiness. We have also identified the operational AI capabilities required to optimise changeover performance. These offer a foundation for developing frameworks that enable manufacturers to navigate AI implementation while maintaining operational efficiency and leveraging lean manufacturing principles, based on the identified ML selection criteria and organisational readiness factors essential for successful AI adoption.

Advanced digital technologies are developed and implemented to address challenges in complex manufacturing systems. Two well-known technologies in this domain are virtual commissioning and discrete-event modeling and simulation. These simulation tools are supported by well-substantiated theoretical frameworks and proven in industrial applications. However, the two approaches target different aspects of production systems and their combination is not extensively researched, thereby motivating this study. This paper reviews existing literature on the combination of these simulation approaches, aiming to identify current opportunities, challenges, and research gaps, while proposing innovative directions for future exploration informed by the findings. The results suggest future research should explore combining VC and DES to enable test-driven development, reduce modeling effort through joint model design, and extend simulation model usability across the production system life cycle.

One of the main challenges in introducing collaborative robots in industry is trust. A commonality in trust definitions in the areas of automation and robotics is the uncertainty associated with unfamiliar situations, where initial trust is influenced by knowledge and experience. For collaborative robots, many safety functions are not visible to users and therefore require knowledge or experience to understand. Becoming familiar with these functions can help users feel less uncertain and vulnerable when working with or near robots. Virtual reality can provide a safe and realistic environment where users can become familiar with new technology. To address this, we propose developing a virtual reality system that enables users to experience some of the safety functions of a collaborative robot. The system is being built in Unity, and the virtual environment includes a representation of an ABB GoFa robot. Through hands-on interactions, users can experience these safety functions and the robot's corresponding behavior when a safety limit is exceeded by using hand-guiding. These events are reinforced using sensory feedback to enhance the perceivability of the safety functions' limits and the robot's safety-related behaviors. For example, vibration intensity increases as the robot approaches its speed limit. The goal of the developed system is to explore whether virtual exposure to these safety functions can help establish initial trust in collaborative robots before real-world interaction, by increasing knowledge and experience.

Collaborative robot (cobot) applications enhance flexibility and efficiency in the manufacturing industry. Even though they are easier to program, their re-programming and transferability across applications remain challenging in fast-changing settings. Artificial Intelligence (AI) technologies reduce the entry barrier to utilising cobots by providing low-code or no-code solutions. This study identifies the requirements for AI-driven no-code solutions for cobot implementation, focusing on technical and managerial perspectives. Through a case study approach informed by the automotive innovation ecosystem, the authors have identified requirements to leverage AI technologies for generating low-code and no-code solutions. These solutions aim to reduce the entry barriers for cobots in manufacturing, enabling agile and adaptive production systems that respond swiftly to market demands. The study highlights the importance of addressing technical and managerial challenges to ensure the successful implementation and value co-creation of cobot applications. 

The emergence of generative AI marks a transformative shift in industrial automation. Traditional robot programming relies on manually written, low-level code that requires specialised expertise, limiting flexibility and accessibility. Recent advances in Large Language Models (LLMs) such as ChatGPT and Mistral introduce new paradigms for automated code generation. However, concerns about data security, model hallucinations, and the opaque reasoning of generative systems continue to hinder their adoption in industry. A promising approach to address these challenges is Retrieval-Augmented Generation (RAG), where the generative model draws on curated, domain-specific data sources controlled by the user. By combining structured knowledge retrieval with generative inference, RAG-based systems can produce robot code that is not only more accurate and context-aware but also verifiable and transparent. This approach enhances user trust and enables safer integration of AI in industrial settings.

This paper explores the application of Retrieval-Augmented Generation (RAG)based architectures - a method that combines information retrieval with LLMs for robot code generation. RAG-based systems enable LLMs to access and utilise domain-specific data, thereby grounding their outputs in reliable knowledge. By leveraging these techniques, robotics developers can achieve more accurate and efficient code generation, potentially accelerating innovation in autonomous systems. Furthermore, it presents a conceptual framework for RAG-enhanced robot programming that balances autonomy with human oversight. The proposed framework enhances the adaptability and intelligence of automated programming by providing a transparent, controllable, and explainable alternative to conventional AI-driven methods, paving the way for more reliable and humancentric automation in future manufacturing environments.

The evolution of industrial automation has moved beyond fixed, repetitive tasks into the realm of intelligent systems capable of adapting to changing production environments. This shift is particularly evident in the transition from Industry 4.0, emphasizing digitalization and cyber-physical systems, to Industry 5.0, which focuses on human-centered, resilient, and sustainable production. At Volvo GTO, a specific assembly station involves the manual removal of protective caps from valve modules-a repetitive, ergonomically straining task that adds little value. Automating this task aligns with both ergonomic improvements and production efficiency goals. The objective of this study is to develop a vision-based robotic solution capable of detecting and removing protective caps with high flexibility and modularity, even in environments with varying geometries and orientations. The system integrates three major components: a YOLOv11-based vision module with 6DoF pose estimation, a custom-designed gripper tool, and a control system implemented in ROS 2 for robot manipulation and task execution.

As the boundaries between human and machine interaction continue to blur, the need for intuitive and efficient interfaces becomes increasingly pressing. In industrial settings, collaborative robots (cobots) have emerged as pivotal tools in facilitating human-machine synergy. This paper presents a comprehensive design, implementation, and evaluation of a voice-controlled cobot interface using open-source software, speech recognition technologies, and large language models (LLMs). The system enables users to issue verbal commands, obtain audio feedback, inquire about process status, and even engage in casual conversation with the robot. By leveraging modular software architecture and locally deployed LLMs, the prototype offers a scalable and user-friendly solution. The study delves into the technical underpinnings, practical limitations, and future possibilities of hands-free robotic control, offering a contribution to the evolving field of human-robot interaction.

Production simulation holds great promise for industrial applications. Virtual commissioning and discrete event simulation are production simulation techniques that are both economically and operationally justified in theory. However, their practical use is limited by high initial costs and expertise, as well as time and effort requirements. This study proposes a model-based engineering framework with a focus on developing and utilizing these two types of techniques in parallel, with the aim of reducing overall costs while simultaneously harnessing the benefits of both methods’ complementary strengths. The proposed framework serves as a basis for future development of methodologies, processes, and tools aimed at streamlining the joint and parallel creation and utilization of said models through simulation-driven systems development. The study presents the framework and an illustrative example that demonstrates its feasibility and practical utility.

Industry 5.0 advocates a human-centric approach where humans play a central role in the design and implementation of industrial technologies. Collaborative robots, with their adjustable safety functions, are key enabling technologies in this paradigm, especially for manual assembly tasks. This study investigates initial trust in collaborative robots by examining whether familiarization through virtual reality (VR) influences the applicability of various trust-related factors. Two groups of university engineering students participated: an Online group that evaluated the factors based solely on images and provided general information, and an In-Person group that engaged in a VR interaction with the robot before evaluation. Participants were asked to assess whether they found factors such as safety, usability, competence, predictability, and adaptability applicable to the robots, selecting "Yes", "No", or "I don’t know". In a follow-up question, they were then asked to rate each factor on a 5-point Likert scale. Results indicate that the In-Person group more frequently affirmed the applicability of the factors. Limitations include the use of only positively framed statements and the participants’ prior experience with industrial and collaborative robots. 

This paper introduces MAXLabs, a distributed cyber-physical testbed designed for advanced manufacturing research. MAXLabs connects geographically dispersed academic production and digitalization labs across Sweden, overcoming localization barriers to create a unified research platform. By leveraging cyber-physical systems, MAXLabs enables remote collaboration, fostering the development, testing, and validation of cutting-edge manufacturing solutions. The initiative aims to enhance global accessibility to these research facilities, promoting innovation in the evolving manufacturing landscape.