主催: (社)計測自動制御学会システムインテグレーション部門
p. 288
Building a robot which acquire body scheme/body image by itself is one of the most intersting issues which aims to build an intelligent robot and understand human intelligence. In the existing work, the robot designer gives the explicit representation of the robot body and needs to re-calibrate it when it has to adapt itself to the changes in its bodies or the environment. In this paper, we propose a method for cross modal map generation which learns constancy of the correlation among multi-modal sensors. The generated map could be regarded as a representation of body surface. A preliminary experiment is shown and future issues are discussed.