Humanoid robot localization based on hybrid map

Abstract

In this paper we present a hybrid map based localization method for humanoid robot. A indoor hybrid map is created by a humanoid robot NAO with a camera and a laser range finder. A global topological map is constructed by natural landmarks, and local metrical maps are established using improved Rao-Blackwellized particle filter. In addition, we set up auxiliary semantic layer based on QR code. and a unified framework of semantic-topological-metric hybrid map is set up. An accurate localization result was obtained by combine a global localization based on topological map and local positioning using KLD-MCL, Experiments shows the method can meet the requirements of indoor localization for a humanoid robot.