A wearable visuo-inertial interface for humanoid robot control

Abstract

This paper describes a wearable visuo-inertial interface for humanoid robot control, which allows a user to control motion of a humanoid robot intuitively. The interface composed of a camera and inertial sensors and estimates body motion of the user: movement (walk), hand motion, and grasping gesture. The body motion (walk) estimation is performed by a combination of a monocular SLAM and a vision-inertial fusion using an extended Kalman filter. The hand motion is also estimated by using the same motion model and sensor fusion as the body motion estimation. The estimated motion was used to operate the movement or arm motion of the humanoid robot. We conducted the experiment on robot operation. The results revealed that the user intuitively controlled the robot and it responded to the operator commands correctly.