Influence of actuator properties on learning to control a virtual limb

Many prosthetic devices incorporate models of muscle dynamics but it is unclear whether this increases learning demands over simpler actuator models. This study determined whether it takes longer to learn to control a virtual limb actuated by a muscle model compared to a simpler force-generator model. Subjects practiced moving a myoelectrically-driven virtual limb from rest to a target position with maximum speed and accuracy. In a muscle model group (n=10), the biceps electromyographic signal activated a virtual muscle that pulled on the virtual limb with a force governed by muscle dynamics, defined by a nonlinear force-length-velocity relation and series-elastic stiffness. A forcegenerator group (n=10) performed the same task, but the actuation force was a linear function of the biceps activation signal. Generalization was assessed with untrained target trials. The results showed that the muscle model group improved performance as fast as the force-generator group and showed greater generalization in early practice. This suggests that incorporating muscle dynamics into a virtual limb does not incur a learning “cost” over the use of simpler actuator models.