3D Indoor Map Building with Monte Carlo Localization in 2D Map

In this paper, we propose a 3D indoor map building method based on Monte Carlo localization in 2D map. The traditional 3D SLAM mainly adopts the visual odometry technology for robot localization. However, the visual localization has a poor real-time performance. Besides, in some special scenarios, such as corridors, the visual localization may generate matching errors, resulting in cumulative errors. These errors will lead to a wrong robot localization. The Monte Carlo localization based on lidar in 2D map can achieve a higher localization accuracy. Therefore, we use the above method to replace the visual localization while using a kinect to collect 3D environment information. To study the performance of the proposed method, we make some experiments and compare with the popular open source RGB-D SLAM system based on visual localization provided by Felix Endres et al. in 2014. The experimental results demonstrate that our method has better effect in the indoor corridor environment with less features.