DistriMuSe - Distributed multi-sensor systems for human safety and health
The DistriMuSe project is dedicated to advancing distributed multi-sensor systems to enhance human safety and health through a seamless Cloud-to-Edge computing architecture. By integrating heterogeneous sensing technologies and artificial intelligence, the project aims to provide robust, real-time monitoring of human behavior and intentions. This technological framework is validated across three critical domains: continuous healthcare monitoring, safety for vulnerable road users, and secure human-robot interaction in industrial environments. Each use case leverages edge-AI to ensure low-latency processing and strict adherence to “Ethics by Design” principles, ensuring both operational efficiency and data privacy.
A significant portion of the project’s research and development is focused on Use Case 3: Human-Robot Collaboration (HRC). In this context, our group leads the development of advanced simulation environments designed to overcome the limitations of traditional data acquisition in industrial settings. We specialize in the generation of synthetic multimodal datasets by simulating complex collaborative robotics scenarios. These datasets include synchronized data from diverse virtual sensors (such as vision, depth, and kinematic sensors), providing a rich and diverse training ground for AI algorithms. This approach allows for the modeling of rare or hazardous edge-case scenarios that are difficult or dangerous to capture in real-world environments, significantly improving the robustness and accuracy of human intention recognition models.
Furthermore, our group is responsible for the implementation of high-fidelity Digital Twins for the industrial scene. These digital counterparts act as an augmented visualization layer, merging real-time sensor data with simulated environments to provide a comprehensive overview of the workspace. By leveraging these Digital Twins, we develop sophisticated collision risk detection algorithms that operate in the continuum between the physical and virtual worlds. The system can predict potential interference between human operators and robotic assets, enabling proactive safety measures—such as real-time path replanning or automated speed reduction—to ensure a safe and efficient co-existence in the factory of the future.
By combining synthetic data generation for AI training with real-time Digital Twin visualization, our contribution ensures that the DistriMuSe framework provides a state-of-the-art solution for industrial safety. This methodology not only accelerates the deployment of collaborative systems but also provides a scalable and secure technical foundation for the next generation of smart manufacturing, where human safety is prioritized through predictive intelligence and advanced digital modeling.
This work has been funded by the project DistriMuSe (HORIZON-KDT-JU-2023-2-RIA 101139769) by the European Union, as well as PCI2024-153511 funded by MICIU/AEI/10.13039/501100011033.
