Virtual muscular fiber force sensing of a musculoskeletal upper limb MultiBody model

Effective therapies are essential in post-stroke re-habilitation to promote full or partial restoration of mobility functions. In particular, accurate estimates of muscle activities and forces that a patient can exert during the movement is of paramount importance to quantify, analyze and monitor the progress of the adopted rehabilitation therapy. Digital twins of musculoskeletal systems provide meaningful and prompt analysis integrating anatomy and physiology characteristics of human beings. Among the numerical techniques, multibody models are largely used in the bio-mechanical community enabling a wide range of studies. Nonetheless, system inputs and operating conditions are often unknown or might differ from reality. State observers, such as Kalman filter techniques, are successfully used in mechanical applications to overcome these limitations linking digital twins and real measurements. In the following contribution, an in-house virtual sensing platform is presented where the multibody model of an upper limb articulation is exploited in a Kalman filter-based scheme to indirectly and jointly estimate muscle forces and marker trajectories during a virtual rehabilitation exercise by means of the well-known OpenSim platform chosen as ground-truth. The contribution presents the virtual sensing framework, the upper limb model description and the estimation results, analyzing the effectiveness and limitations of the proposed approach.