Inferred control patterns depend on model complexity

To see the implications of model complexity for the inverse problem of identifying control signals, four models of wrist flexion were simulated. The most complicated model consisted of an inertial load driven by two Hill-type models, with the exception that series elasticities were separated into activation-dependent (contractile element) and activation-independent (tendon) parts. The simplest consisted of an inertia driven by a net torque. Model outputs are compared to actual wrist flexions, assuming simple pulse-shaped inputs, with two pulses per muscle. Pulse heights, pulse widths, and the delay between pulses were varied to achieve best fits. The best fit control signals are compared to recorded electromyograms