Empirical study of future image prediction for image-based mobile robot navigation

Recent image-based robotic systems use predicted future state images to control robots. Therefore, the prediction accuracy of image affects performance robot. To predict images, most previous studies assume that camera captures entire scene and environment is static. However, in real robot applications, these assumptions do not always hold. For example, if a attached mobile robot, its view changes from time time. In this study, we analyzed relationship between model robot’s behavior, controlled by an navigation algorithm. Through experiments using front-faced omni-directional cameras, discussed capabilities models demonstrated their when applied Moreover, adapt dynamic environment, studied effectiveness directing ceiling. We showed robust can be achieved without cameras directed toward front or floor, because views disturbed moving objects environment. • Empirical evaluation comparison on Action-conditioned LSTM-based performs well across types. Omni-directional effective for models. tasks.