Identification of a Step-And-Brake Controller of a Human Based on Prediction of Capturability.

An explicit mathematical form of a human's step-and-brake controller is identified through motion measurement of the human subject. The controller was originally designed for biped robots based on the reduced-order dynamics and the model predictive control scheme with the terminal capturability condition, and is compatible with both stand-still and stepping motions. The minimal number of parameters facilitates the identification from measured trajectories of the center of mass and the zero-moment point of the human subject. In spite of the minimality, the result only suited the human's behaviors well with slight modifications of the model by taking direction-dependency of the natural falling speed and the inertial torque about the center of mass into account. Furthermore, the parameters are successfully identified even from the first half of motion sequence, which means that the proposed method is available in designing on-the-fly systems to evaluate balancing abilities of humans and to assist balances of humans in walking.

Identification of COM Controller of a Human in Stance Based on Motion Measurement and Phase-Space Analysis.

This article proposes a process to identify the standing stabilizer, namely, the controller in humans to keep upright posture stable against perturbations. We model the controller as a piecewise-linear feedback system, where the state of the center of mass (COM) is regulated by coordinating the whole body so as to locate the zero-moment point (ZMP) at the desired position. This was developed for humanoid robots and is possibly able to elaborate the fundamental control scheme used by humans to stabilize themselves. Difficulties lie on how to collect motion trajectories in a wide area of the state space for reliable identification and how to identify the piecewise-affine dynamical system. For the former problem, a motion measurement protocol is devised based on the theoretical phase portrait of the system. Regarding the latter problem, some clustering techniques including K-means method and EM (Expectation-and-Maximization) algorithm were examined. We found that a modified K-means method produced the most accurate result in this study. The method was applied to the identification of a lateral standing controller of a human subject. The result of the identification quantitatively supported a hypothesis that the COM-ZMP regulator reasonably models the human's controller when deviations of the angular momentum about the COM are limited.