Abstract
In this paper, we propose a method to estimate muscle activity from joint angle data. We use a model which consists of elastic bodies as muscles and rigid bodies as bones and utilize finite element method and robotics technique. This method can consider change of muscle force accompanied by change of muscle fiber direction. Muscle activity is estimated by finding activity to generate the reference joint torque which is calculated by inverse dynamics. The estimation is executed as sequential quadratic programming and can be solved in parallel computation for each muscle. We compared execution time of parallel computation and sequential one, and estimated muscle activity. As a result, the efficiency of parallel computing was indicated and the muscle activity which generates reference joint torque was obtained.
