Estimation of Hand Position and Posture using Inertial Sensors and its Application to Robot Teaching System

Japanese society is rapidly aging, and the birthrate is low, leading to a reduction of the working population. To compensate for the resulting labor shortage, particularly in the field of teaching and training, robots can be considered. Conventional teaching methods with robots include remote teaching and direct teaching. These teaching methods have problems in terms of difficulty teaching sensibly and teacher safety. We extracted the target position and posture of a robot to be taught using motion sensors, and its performance was verified by simulation before applying our system to the actual robot. As a result of comparing our system to optical motion capture, we also found that it is unstable to estimate the initial posture angles around the z-axis by the geomagnetic sensor, and it is necessary to correct these angles in order to improve the position and posture estimation accuracy. In this system, the error of the sensor mounting position influences the deviation of the estimated trajectory in direction to the z-axis. As a result of the simulation, we found that the hand trajectory estimated from the motion sensor can be applied as the endpoint trajectory of the manipulator.