Preliminary Validation of Upper Limb Musculoskeletal Model using Static Optimization

Musculoskeletal models are powerful analogues to simulate human motion through kinematic and dynamic analysis. When coupled with feature-rich software, musculoskeletal models form an attractive platform for the integration of machine learning for human motion analysis. Performing realistic simulations using these models provide an avenue to overcome constraints when collecting real-world data sets. This motivates the need to further investigate the validity, efficacy, and accuracy of each available model to ensure that the resultant simulations are transferable to real-world applications. Using the open-source software, OpenSim, the primary aim of this paper is to validate an upper limb musculoskeletal model widely used in research. Muscle activation results from static optimization are evaluated against real-world data. A secondary aim is to investigate the effects of two muscle force generation constraints when evaluating the model's validity. Results show an agreement between the optimized muscle activation trends and real-world sEMG readings. However, it was found that static optimization of the musculoskeletal model is unable to identify voluntary co-contractions since the redundant model has more muscles than the system's degrees of freedom. Thus, future work will look to utilize additional channels of information to incorporate this during analysis.