Towards Scalable and Context-Aware Digital Twins for HRC

Digital twins are virtual representations of physical systems that enable real-time simulation, monitoring, and optimization. By creating a shared digital space that mirrors the real world, they facilitate better decision-making, predictive maintenance, and operational efficiency. As a core technology in many industries, digital twins offer the potential to enhance safety, flexibility, and productivity in complex systems. Their adoption is particularly promising in Human-Robot Collaboration (HRC), where humans and robots work together in close proximity. Digital twins synchronize physical and virtual systems, allowing tasks to be planned, monitored, and executed with precision while maintaining operator safety. However, traditional digital twin implementations often face challenges related to scalability, modularity, and state management, especially in dynamic, real-time HRC scenarios involving a high number of agents. These limitations reduce their ability to support large-scale systems, adapt to diverse applications, and provide robust analytical capabilities. This thesis focuses on developing a modular digital twin architecture that is versatile and easy to adapt across a wide range of use cases with minimal effort. The proposed framework emphasizes modularity, enabling seamless extensibility by allowing new robots, models, and logic units to be added without the need for reprogramming or redeployment. This significantly reduces development time and complexity while facilitating the creation of new workspaces tailored to specific needs. Additionally, recognizing the critical role of state history in digital twins for analyzing past system behavior and enhancing predictive capabilities, the architecture incorporates an efficient and robust state history system. This feature ensures precise and consistent tracking of changes over time, empowering users with deeper insights for decision-making and optimization in both real-world applications and simulated environments. The aforementioned architecture has been validated across multiple use cases in Human-Robot Collaboration (HRC) scenarios, including bi-directional communication between human operators and different kinds of robots, kinesthetic teaching throughout mixed reality, and benchmarking robotic plans. To ensure its practicality and effectiveness, the architecture has been benchmarked against other digital twin systems in the literature in a complex setup, with a particular focus on critical metrics such as CPU usage, network traffic, latency, real-time factor, and dropped message. The results demonstrate stable performance with real-time factor of 1.0 while maintaining latency lower than 7ms and predictable CPU usage, highlighting the architecture’s ability to handle complex, dynamic environments while maintaining a comprehensive state history.