Improving physiological fidelity of in vitro knee simulators through bidirectional optimized muscle control

Physiological in vitro knee simulators enable invasive biomechanical evaluation of surgical interventions; however, most existing simulator controllers are limited to single-direction squatting activity with a restricted range of motion, and lack active agonist-antagonist co-activation, thereby reducing physiological fidelity. This study developed a generalizable control framework integrating bidirectional actuation and force optimization to enable testing across multiple activities of daily living. The BiOME knee simulator implemented bidirectional sagittal-plane control through coordinated actuation of inferosuperior and anteroposterior ankle translation. The control architecture combined feedback position and force loops with a model-based feedforward estimator and an optimisation routine to generate physiologically-relevant quadriceps-hamstring co-contraction. Five control strategies combining these elements were implemented for squatting motion on a phantom knee and benchmarked against published in vivo and in vitro data. The inclusion of the feedforward block improved vertical ground reaction force (GRF) tracking accuracy by 48 % and flexion-angle accuracy by 80 % with high repeatability. indicating that incorporating model-based compensation further enhances system accuracy. Optimised bidirectional control restored ankle dorsiflexion to physiological ranges, reduced peak quadriceps loads by 37 % producing closer agreement with the literature. Unlike existing simulators requiring activity-specific hardware redesign, this generalizable framework requires only force plate and motion capture data from the desired activity to generate physiologically appropriate agonist-antagonist co-contraction, enabling replication of diverse activities such as squatting, cycling, kneeling, and stair climbing for testing implants and surgical techniques.