Design and development of a postural prediction system for Digital Human Model in 3D virtual work-station assessments

Awareness of how important it is to preserve employees' physical and mental health has grown in the past decades. Creating a safe and ergonomic work environment has become one of the main concerns of engineers and researchers nowadays. The integration of new technologies into the workplace, such as artificial intelligence and virtual and augmented reality, to design a human-centered work environment is investigated continuously. Virtual reality is the most widely used technique in the latest ergonomic assessment approaches in which human-workstation interaction can be simulated using digital human models (DHM) while the workstation can be imported using CAD models. The applications developed so far are quite sophisticated and need extensive training. In this frame, the proposed thesis aims to implement a simple-to-use, low-training system for the ergonomist, in which the DHM is animated by inverse kinematics. The postural prediction system predicts the behavior of a DHM with different anthropometric percentiles within a virtual workstation. To implement the model, the Unity3D engine and an inverse kinematic tool, the Final IK, were used. To predict an ergonomic posture the Final IK was supported by postural rules. Xsens Awinda technology was used to record a motion capture session and evaluate the input parameters for the postural prediction system. The user replicates a work task, on-site or in a virtual environment. Hands and feet positions and rotations, as well as head rotations, are extracted. Alternatively, if a motion capture session is not feasible, the user can insert the target's coordinates or move them via keyboard commands. The user has the opportunity to move the end effectors (hands and feet) but also the hips, shoulders, and pelvis. This is useful either to assess a specific posture or to fine-tune the posture erroneously predicted by the system. If a virtual workstation is loaded into the Unity scene, the system is also able to detect possible collisions with objects and use this information to adjust the predicted posture accordingly. The testing phase involved the recording of several motion capture sessions. The user was always different in each session to represent different percentiles and highlight critical issues in the prediction system. The experimental phase confirmed the advantages of introducing postural rules and constraints imposed by the workstation to obtain more realistic postures. This project confirms what emerged from the related literature review: these applications are useful to highlight any technical issues caused by workstation organization, before producing any physical prototypes. Thus, allowing for faster intervention in the workstation design while saving time, cost, and resources.