Abstract
In this study, we conduct experiments of human balancing task and present a stochastic dynamic model of the human balancing task. The parameters of the stochastic dynamic model are identified using particle swarm optimizers. Tn order to examine stochastic behavior of the human balancing, we experimentally estimate probability density function of the state quantities. The experimental result implies that the behavior of the human balance seems to be fat-tailed distribution. This experimental observation was reproduced by the model that has additive and multiplicative white Gaussian noise. This work has implications for designing human-like motions of artificial agents such as human-like partner robots.
