An improved muscle-reflex actuator for use in large-scale neuro-musculoskeletal models.

This paper extends the systematic approach described in Winters and Stark (62) for developing muscle models. The underlying motivation is our finding that for larger scale shoulder and head-neck postural systems to be mechanically stable, open-loop muscle properties are often not sufficient. There are three primary contributions. First, the previous muscle mechanical model structure and parameter estimation process of (62) is updated to reflect recent experimental findings. Second, an intrafusal (IF) muscle model is developed that includes a gamma static motoneuron (MN) drive, a Hill muscle model, and a muscle spindle sensor across the IF series element; this provides a more appropriate muscle spindle output signal, especially for studies of posture. Third, the conceptual cut between the neuro-control input and the actuator is raised from just below the MN summing junction to a higher location, allowing a "muscle-reflex actuator" to be defined that satisfies the formal theoretical requirement for possessing passive spring-like behavior when the neurocontrol input is constant, alpha-gamma MN coactivation is assumed, and three types of intrinsic autogenic reflex responses (spindle, Golgi tendon organ, Rhenshaw cell) are developed. Default feedback gains are set based on the criteria that inherent feedback should not sculpt the feedforward excitation drive by more than +/- 10% of maximum. This new actuator model only mildly affects voluntary goal-directed dynamic performance, but enhances spring-like performance around the postural equilibrium state, in line with available animal and human studies and with several theories on postural regulation.