Rapid identification of muscle activation profiles via optimization and smooth profile patches

The identification of muscle activation profiles for general motion tasks of the musculoskeletal system is an important and still open problem in biomechanics. In the context of simulation of surgical interventions (e.g. tendon transfer [4] or the shortening/lengthening of specific muscles that significantly influence a pathological motion [5]) the knowledge of the contribution of a single muscle to joint moments during motion can yield information that enables the physician to directly assess a therapy before applying it to a patient. Besides EMG‐driven models, which rely on measured muscles activity and static optimization ([6] and [7], respectively), current methodology consists of dynamic optimization where the amplitude and the time history of the activation function is sought by forward dynamic simulations. In this setting, function parameterizations such as discretization through sampling values, rectangular and ramp functions, as well as splines have been investigated. However, typical optimization runs display prohibitively long computation times, which make them unsuitable for on‐line approximation of dynamical motions, such as required for example for patient‐specific therapy. The present paper proposes a new method for parameterizing muscle activation profiles employing smooth (C^Ë) functions. We analyze the method in comparison to an input sampling approach for a flexion/extension two‐joint subsystem of hip and knee, comprising two pairs of antagonistic muscles. The results show that, by using smooth base functions for muscle activation, the number of design variables as well as the required CPU time can be reduced without loss of quality of the ensuing simulations. The method is currently under investigation and is planned to be applied to a full‐muscle system of human gait in the near future.