Conference Agenda

Natural and intuitive human–robot interaction is increasingly important as robotic systems are progressively integrated into human environments. Conventional teleoperation approaches often rely on dedicated interfaces or wearable sensors, which increase system complexity and limit accessibility.

This work proposes a vision-based framework for robot teleoperation that exploits a human digital twin (HDT) as an intermediate layer between motion capture and robot control. Human arm motion is acquired using an RGB-D camera (Luxonis OAK-D Pro) combined with a YOLOv8-based pose estimation pipeline, enabling markerless tracking of upper-limb joints. The extracted trajectories are processed in MATLAB/Simscape to drive a multibody human model via inverse kinematics, generating biomechanically consistent motion. The resulting wrist trajectory is then used as input for the inverse kinematics of a robotic manipulator.

Unlike conventional leader–follower schemes, the proposed architecture introduces a digital twin layer that enables visualization, validation, and filtering of human motion prior to execution. A pick-and-place experiment is conducted to evaluate the framework and assess the influence of camera placement on tracking accuracy. Results show that an oblique camera configuration significantly improves performance compared to frontal and top-down views, highlighting the importance of sensing geometry in vision-based teleoperation systems.

Accurate estimation of joint centres from skin-mounted marker trajectories is a prerequisite for anatomically plausible virtual human animation in biomechanics, ergonomics and computer graphics. Marker-based stereophotogrammetric systems provide high-resolution kinematic data, yet the inherent offset between epidermal markers and the underlying skeletal structures imposes a fundamental limitation on direct retargeting approaches. This paper presents a ge-ometric reconstruction framework that infers joint centre locations and skeletal segment axes from Vicon optical motion-capture data without assuming a one-to-one correspondence between external landmarks and internal anatomical nodes. The method employs a three-sphere tangent-line construction applied to selected marker subsets to estimate the proximal and distal articulations of each rigid segment. Applied to the femoral segment, the algorithm was validated on three kinematically distinct motor tasks—straight walking, vertical jumping and running—obtained from exhaustive combinatorial sampling of all three-frame subsets. The median femoral length remained remarkably consistent across tasks (0.491, 0.486 and 0.471 m; cross-task range 21 mm), and the optimised sphere radii were anatomically coherent with published marker-to-bone distances. A sensitivity analysis identified the pelvic-midpoint radius as the dominant source of length variability (|r| > 0.95), while outlier triads were traced to geometrically degenerate frame selections with insufficient postural diversity. The proposed framework is directly applicable to subject-specific biomechanical analyses, musculoskeletal modelling and movement simulations where accurate, non-invasive joint centre estimation from surface markers is required.

Brain atlas selection is rarely treated as an explicit methodological variable in neuroimaging workflows, yet successive releases of probabilistic atlases may differ substantially in their spatial coverage and ontological completeness. This paper examines how the anatomical assignment of a glioblastoma (GBM) lesion changes across five successive versions of the Julich Brain Atlas (v1.18, v2.4, v2.6, v3.0, v3.1) as deployed on the EBRAINS research infrastructure, within the framework of the EU Horizon project DTRIP4H. A single IDH-wildtype GBM case - segmented via MONAI from clinical DICOM data and registered to the MNI ICBM152 reference space - was mapped against each at-las version using labeled maximum probability maps. Volumetric intersection statistics were computed separately for enhancing tumor and necrotic core, without diagnostic or functional inference. Results show that atlas coverage of the tumor volume ranges from 49.8% to 82.4% across versions, with no mono-tonic relationship to release recency. Ontological labeling quality varies inde-pendently, with three versions unsuitable for named-area assignment in their current deployment. Version v3.0 provides the most complete and interpreta-ble output. A potential anatomical discontinuity between v3.0 and v3.1 is doc-umented and discussed.

The development of Autonomous Surface Vehicles (ASVs) requires simulation tools capable of supporting rapid prototyping, perception payload integration, and repeatable scenario-based testing before field deployment. In this context, the paper presents a modular ROS2/Gazebo virtual testbed for ASV simulation, with particular focus on conditional awareness payloads and configurable marine scenarios. The proposed testbed is organized around reusable vehicle, payload, environment, and interface modules, allowing different ASV configurations to be assembled without redefining the complete simulation model. A common ROS2 interface layer enables Software-in-the-Loop execution and provides the basis for Hardware-in-the-Loop workflows. The testbed integrates configurable wave and wind environments, virtual radar and LiDAR payloads, and a scenario-management layer for static and dynamic obstacle interaction. Two representative ASV configurations are considered: a baseline navigation setup and a conditional awareness setup with perception payloads and obstacle management. A preliminary runtime evaluation shows the computational impact of progressively enabling payload and scenario modules, while maintaining near real-time execution in the evaluated conditions. Future work will address validation against field data and sea trials.

This study presents the development and experimental validation of a 3D Finite Element Method (FEM) Digital Twin to analyze the mechanical response of submarine optical fiber cables used for Distributed Acoustic Sensing (DAS). Focusing on the FOCUS infrastructure deployed in the Ionian Sea, the research investigates the strain transfer mechanisms of both Tight Buffered and Loose fiber architectures. The numerical submodels, subjected to kinematic control and Zero-Strain Boundary Conditions, successfully replicate the macroscopic phenomena observed through Swept-Wavelength Interferometry. For the Tight configuration, the Digital Twin accurately reproduces a homogeneous longitu-dinal strain plateau. For the complex Loose configuration, a phenomenological approach utilizing non-linear frictional contact and radial offset was imple-mented to overcome numerical instability. The validated model successfully captures the non-linear shear-lag effect and the highly asymmetrical strain transfer between tension and compression phases, analytically demonstrating the critical role of Excess Fiber Length and micro-buckling in activating Cou-lomb friction. This validated Digital Twin provides a robust baseline for pre-dicting cable performance under complex environmental loads.