3D self-localization for humanoid robots using view regression and odometry

Abstract

The ambience of research in VSLAM and relocalization algorithms in the last few years allure with real-time localization and increased precision for RGBD or stereo cameras but with ambivalence requesting higher computational capacity or more expensive sensors. The aim of this paper is to present a gentle algorithm to locate a humanoid robot using relocalization view based algorithms and odometry information using general regression with Nadayara-Watson kernel for applications where the area is already known such as RoboCup competitions or service robots.