Digital Twin of the FloatEvo Upper Limb Exoskeleton in MuJoCo: Design, Sensor Integration and Control

In recent years, the development of digital twins for robotic and rehabilitation systems has gained increasing attention due to their ability to provide accurate virtual replicas of physical devices for analysis, testing, and control. This thesis presents the design of a digital twin of the FloatEVO upper-limb exoskeleton, developed at the Rehab Technologies Lab of the Italian Institute of Technology (IIT), using the Multi-Joint dynamics with contact (MuJoCo) simulation environment. FloatEVO is a 7-Degrees of freedom (DoF) robotic exoskeleton designed for upper-limb rehabilitation and assistive therapy, targeting both the scapular and glenohumeral complexes as well as the elbow joint. Its structure reproduces the main articulations of the human arm, allowing for natural range of motion and for the possibility of implementing multiple control modes, including assist-as-needed rehabilitation exercises. The system aims to support patients with motor impairments by providing intensive, repeatable, and measurable therapy sessions, enhancing both recovery effectiveness and clinical assessment. The objective of this thesis is to develop a high-fidelity digital counterpart of FloatEVO, that should be able to replicate its mechanical structure, kinematics, and sensorization. The model reproduces the complete kinematic chain of the exoskeleton, including the scapular, shoulder complex and elbow joints, with each joint equipped with virtual torque sensors for monitoring joint torques and interaction dynamics during simulated movements. Several simulations were conducted to validate the accuracy and dynamic consistency of the model with respect to the physical system, such as the estimation of joint torques in different configurations, the reproduction of recorded movements from experimental data, and the computation of the shoulder joint center (glenohumeral center) during dynamic motion. Furthermore, a real-time streaming framework was implemented to synchronize joint position data from the physical exoskeleton to its digital twin, allowing the virtual model to mirror the real robot’s movements and estimating biomechanical parameters. The choice of the MuJoCo simulation environment was motivated by its ability to perform both direct and inverse dynamic simulations with high numerical stability and computational efficiency. Beyond its technical significance, the creation of the digital twin has strong clinical motivations for supporting rehabilitation of patients with orthopedic and neurological impairments, including post-stroke conditions, traumatic injuries, and surgical interventions affecting shoulder mobility. It provides a safe and versatile environment for replicating rehabilitation sessions, monitoring the patient’s kinematic behavior, and, in perspective, simulating personalized rehabilitation protocols tailored to each individual’s motor performance. This approach opens the possibility of quantitatively assessing recovery progress, optimizing therapy design, and integrating predictive modeling into patient-specific rehabilitation planning. Ultimately, the developed framework establishes a foundation for real-time digital mirroring between FloatEVO and its virtual counterpart, supporting future advancements in adaptive control, sensor integration, and clinical assessment. The developed digital twin thus serves both as a research platform for testing and validation, and as a potential clinical tool to enhance rehabilitation outcomes through personalized and data-driven approaches.