PREDICTION OF 90-DEGREE TURN USING KINEMATIC PARAMETERS FOR IMPLEMENTATION OF HOME ROBOT

Presently, social trends are indicating that single elderly households are increasing while lifestyles are diversifying. This paper proposes the Biofied building concept. In Biofied buildings, home robots provide a safe, secure, and comfortable space for residents. The objective of introducing robots is to sense a resident’s position and movement in real-time. However, with the existing system, it is difficult to prevent unexpected accidents and control robots without disturbing the residents. The objective of this study was to incorporate the motion prediction phase into the system and realize more natural spatial control using robots. Although a simple motion prediction method based on dynamics using system identification has been proposed by a previous study, this cannot be applied to non-steady motion. Hence, it is necessary to extend the system such that it can be applied to various behaviors observed on a daily basis because the nonlinear behavior is not versatile. Particularly, with regard to walking motion, it has been reported that more than 40% of walking motions in daily life are non-steady, and the 90-degree turn accounts for approximately 20% of non-steady motions. Thus, the 90-degree turn was selected as the prediction target, because it is a basic non-steady walking motion and the most essential motion in daily life. This study investigated the effectiveness of IMU sensors in measuring movements with complicated posture control. Notably, these sensors are inexpensive and not restricted by the measurement environment; therefore, the proposed method was developed based on data obtained using IMU sensors. The kinematic parameters extracted in the preliminary experiment are useful for predicting the 90-degree turn by comparing steady walking to the 90-degree turn. Through comprehensive observations, the yaw and roll of the head and upper trunk, and the acceleration of shoulders and shanks, were extracted as the prediction parameters. The 90-degree turn, which was performed at the subject’s own time and according to their free will, was measured to validate the prediction of parameters under the condition wherein the environmental and mental effects are eliminated. The prediction method was developed using the threshold defined by the amplitude during steady walking. The proposed method was evaluated by considering the prediction accuracy and prediction time. The fastest prediction time of 0.85 seconds was that of the yaw angular velocity at the