Abstract
A neuro-musculoskeletal model which is similar to the actual man was developed in order to predict a recovery level of pathological gait and to restore locomotion of human ancestor. The musculoskeletal system was modeled as two dimensional nine rigid links and twenty-two muscles in the sagittal plane, and as two rigid links and two muscles in the horizontal plane. The neural system was modeled as a rhythm generator composed of fourteen neural oscillators. Genetic algorithms were used to determine those neural parameters. Bipedal walking was synthesized as mutual entrainment between the rhythmic activities of body dynamics and the oscillation of neural system. Criterion for the parameter searching was defined firstly as walking distance and the number of steps until falling down. After a stable walking was achieved, then the parameters were adjusted to agree with the measured walking, in terms of stride length, walking cycle, and joint moments. The synthesized walking with an actual human body proportion was in good agreement with the actual walking. As examples of different body proportions, infant and ape models were simulated with minimum energy criterion instead of the criterion of similarity to actual walking. They were also agreed to those of actual walking. These results show that walking pattern corresponding to body shape could be synthesized by the proposed method.
