Toward developing bouncing-type model for simulating human running behavior

(Improvement of identification accuracy by nondimensionalization)

Abstract

The passive running is a method to descend a slope by a running gait only using potential energy. We have already proposed an extremely simple passive running model named bilateral arcs bouncing (BAB) model that consists of two arcs which face each other. In this study, we improve our previous parameter identification method for our BAB model that reproduces human running behavior. We conducted an experiment to obtain time series data of the body trunk movement during the human running behavior. Based on the obtained data, we solved a parameter identification problem to fit our BAB solutions to the experimental data, by proposing a modified particle swarm optimization and also deriving nondimensionalized equations of motion. We employ the cost function based on maximum absolute error between the phase orbits of the human running behavior and our BAB behavior. The resulting model yields over 95% accuracy of reproduction of the phase orbits of the human subjects during running behavior.