Medical Mobile Robot Localization in Hospital Corridor Environment Using Laser SLAM and Text Features

Abstract

Localization is an important capability for a medical service robot navigating in the hospital environment. Recently, by using laser-based simultaneous localization and mapping (SLAM) and Monte Carlo localization (MCL) techniques, the mobile robot can localize itself in those places with obvious geometric features, like corners, lines, breakpoints, and curves. However, when a robot moves in a long corridor environment inside a building, there are only a few geometric features for reference, especially in symmetrical or similar areas. Consequently, the mobile robot can neither localize itself correctly in the global localization phase without a known initial pose nor recover the correct pose from the kidnapped robot problem. To solve this problem, we propose an approach that uses text features on the doorplate to assist localization. In the mapping stage, we control the mobile robot moves along a fixed path using both text features and laser scans data. In the localization stage, the mobile robot moving according to a strategy to find text features and initializes the particles, then navigates and localizes using an improved MCL method. The experimental results show that our approach achieves over 90% mapping accuracy and 93.3% successful localization rate, which outperforms traditional methods.