Christian Eichhorn

Clinical workflows are inherently complex, and evaluating future technologies, e.g., the involvement of a humanoid robot, with patients is challenging. A bridge is needed when transitioning from the laboratory to the hospital. Immersive Virtual Environments (IVEs), a proven strategy for in-silico training scenarios, can ensure integration of technologies into existing care workflows, without the need for physical patients. We developed an authentic IVE in Unity that mimics a specific European hospital, including lifelike patient agents driven by a Large Language Model (LLM). Based on the findings of a pilot study, the focus was placed on including trigger-based physical props and integrating multi-user functionality. In a mixed 2x2 study design, 32 clinic staff performed an in-silico wound care scenario, providing care to the LLM-based patient. This main study is divided into two parts: Study I - the use of physical props vs. fully virtual objects; Study II - the collaborative work with another person (multi-user) vs. performing the task alone. We measured usability and presence to evaluate the physical props and the multi-user experience. These real-world results have been further analyzed, focusing on pre-existing knowledge in wound care management. Lastly, we compare the impact of the individual features overall on the simulation. Findings are then placed in context using existing studies from related publications. Our findings include: physical props contribute to presence and, through familiarity, also provide an entry point to the simulation. However, a personal selection of items as well as high fidelity is essential. Multi-user enhances presence but adds complexity to achieve the same level of task fidelity, suggesting the need for an initial training session. Other remarks include a promising future perspective on LLM-based patients, with an emphasis on situational awareness (following the action), for which more research will be necessary.

Conventional Augmented Reality (AR) head-mounted displays (HMDs) integrate all essential components within the wearable unit, imposing limitations on weight, heat dissipation, and battery life. Beaming Displays (BDs) address these constraints by externalizing computation and projection; however, achieving reliable low-latency six-degrees-of-freedom (6-DoF) tracking remains a challenge. We present BeamStellar, a BD system that enables low-latency 6-DoF head tracking by combining a position-sensitive detector (PSD) with a spatio-temporally modulated infrared LED array mounted on passive glasses. The system integrates tracking and projection via a co-aligned optical path and performs real-time signal processing on an FPGA to minimize tracking latency. We describe the system architecture, pose reconstruction pipeline, and evaluate the pose-tracking latency of approximately 430 μs. BeamStellar provides a replicable, lightweight AR tracking platform, establishing a new paradigm for low-latency environment-based AR systems.

This systematic literature review analyzes the role of dynamic difficulty adaptation (DDA) in serious games (SGs) to provide an overview of current trends and identify research gaps. The purpose of the study is to contextualize how DDA is being employed in SGs to enhance their learning outcomes, effectiveness, and game enjoyment. The review included studies published over the past five years that implemented specific DDA methods within SGs. Publications were identified through Google Scholar (searched up to 10 November 2025) and screened for relevance, resulting in 75 relevant papers. No formal risk-of-bias assessment was conducted. These studies were analyzed by publication year, source, application domain, DDA type, and effectiveness. The results indicate a growing interest in adaptive SGs across domains, including rehabilitation and education, with DDA methods ranging from rule-based (e.g., fuzzy logic) and player modeling (using performance, physiological, or emotional metrics) to various machine learning techniques (reinforcement learning, genetic algorithms, neural networks). Newly emerging trends, such as the integration of generative artificial intelligence for DDA, were also identified. Evidence suggests that DDA can enhance learning outcomes and game experience, although study differences, limited evaluation metrics, and unexplored opportunities for adaptive SGs highlight the need for further research.