抄録
Neuro-musculo-skeletal model has been developed that is capable of generating reasonable gait. However, the details of the ground reaction force waveform differ from real gait motion quantitatively. A whole-body walking model including the detailed foot model has not been proposed and a unified objective function has not been proposed. Therefore, the purposes of this study are to improve a forward dynamic walking simulation by developing a precise foot model and making a hybrid criterion for the generation of human gait motion by using the application and utilization of genetic algorithm for the optimized calculation. The dynamics of the human body has been represented by 22 three-dimensional rigid links with 78 muscles and 62 neural oscillators. Unknown neuronal parameters have been adjusted by a numerical search method using the evaluative criterion for locomotion that is defined by the locomotive energy efficiency, the toe clearance and the magnitude of the ground reaction force. As a result of adjustment of the neuronal parameters, walking 10 steps has been generated. The simulated walking pattern has closely agreed with actual human walking not only of joint movement but also of the mutual conversion of the mechanical energy and the magnitude of the ground reaction force.
