A 2-D Indoor Localization System Using 3-D Structural Features for Mobile Robots

Indoor environments, as typical unstructured settings, present significant challenges for the localization of mobile robots. In such environments, robots are prone to getting lost or mismatched to incorrect locations. Existing solutions heavily rely on 2-D light detection and ranging (LiDAR), which can only scan the horizontal plane of the environment, thus failing to fully observe spatial objects, resulting in insufficient available features for the localization of robots. In response to this challenge, this article introduces a system that measures 3-D structural features for 2-D localization. First, a vision sensor is employed to capture the 3-D structural features of the scene. A hierarchical strategy is then introduced to extract key structural features, mapping the 3-D features into 2-D hierarchical submaps. A map selection algorithm is further proposed to filter the localization map. Next, we propose a method to convert point cloud data into 2-D pseudo-laser representations, allowing for parallel matching between the hierarchical submaps and the pseudo-laser data to obtain multiple localization results. Building on this, we investigate an observation residual evaluation method to assess the performance of multiple localization results, enabling fused localization. Both simulation and real-world experiments demonstrate that the introduced approach significantly improves the accuracy and robustness of localization for mobile robots.