Text-MCL: Autonomous Mobile Robot Localization in Similar Environment Using Text-Level Semantic Information

Localization is one of the most important issues in mobile robotics, especially when an autonomous robot performs a navigation task. The current and popular occupancy grid map, based on 2D LiDar simultaneous localization mapping (SLAM), suitable easy for path planning, adaptive Monte Carlo (AMCL) method can realize rooms indoor environments. However, conventional fails to locate there are similar repeated geometric structures, like long corridors. To solve this problem, we present Text-MCL, new text information laser scan data. A coarse-to-fine paradigm used localization: firstly, find coarse place global by finding text-level semantic information, then get fine local using (MCL) Extensive experiments demonstrate that our approach improves speed success rate 96.2% with few particles. In addition, proposed recover from kidnapping after short movement, while MCL methods converge wrong position.