Research on the virtual space system used for learning of autonomous robots

Abstract

Recently, the applications of the virtual space become more popular. In some cases, the experiments in real space require huge time and huge costs. For such cases, experiments in the virtual space are applied. We have developed virtual space system for learning of autonomous robot controlled with visual information in OpenGL space. The purpose of this study is to develop the more realistic virtual space which can compensate the difference between real space and virtual space. For this purpose, we introduced the texture mapping to the object in virtual space and the additional virtual sensors which can use for controlling the robot. We made some experiments of autonomous robots controlled by SOM-RL algorithm and examine the effectiveness of proposed system.