How It Works

NCP takes in the Synergy Extrapolation results derived from EMG data, and it reoptimizes the constant synergy activation weights in order to reduce error according to the cost function. It iterates through the motion and finds the synergy values that most closely match the experimental muscle activation values and the modeled joint moments.

It also estimates muscle synergy sets for areas of the body that had no EMG data to begin with. It does this by running an optimization to stay as close as possible to the experimental movement data with a few additional built in assumptions to make up for the lack of EMG data. For example, in a walking simulation, it may assume the same activation levels for the legs and the trunk in order to fill in synergy information for the trunk.

MTLI can be used in conjunction with NCP in order to improve accuracy by setting initial values for optimal fiber length, tendon slack length, and maximum isometric force of muscles before running the NCP optimization.

Optimizing Muscle Synergies

Neural Control Personalization uses a few cost terms to find a unique set of muscle synergies that fulfill certain requirements.

• moment_tracking - experimental joint moments from inverse dynamics should be tracked closely
• activation_tracking - experimental EMG-derived activations (from MTP) should be tracked closely
• activation_minimization - non-EMG derived activations should be minimized to prevent excessive co-contraction. This is a tunable cost term depending on the movement.
• grouped_activations - muscle activations in the same muscle group should be similar (see activation groups)
• grouped_fiber_lengths - muscles in the same fiber length group should have similar fiber lengths relative to their optimal fiber length (see fiber length groups)
• bilateral_symmetry - muscles should have similar synergy weights to their contralateral counterparts