PINMAP: A Cost-Efficient Algorithm for Glass Detection and Mapping Using Low-Cost 2-D LiDAR

Autonomous mobile robots (AMRs) have seen rapid adoption due to their ability to autonomously navigate, avoid obstacles, and collaborate efficiently in complex environments. AMRs equipped with light detection and ranging (LiDAR)-based simultaneous localization and mapping (SLAM) are effective in obstacle-rich settings. However, SLAM approaches, particularly those using low-cost 2-D LiDAR, face challenges in accurately detecting and mapping glass surfaces. AMRs may interpret glass as open space, potentially leading to collisions. In this article, we propose a novel framework, the probabilistic incremental navigation-based mapping with accumulative point cloud data (PINMAP), which enables glass detection and mapping without additional sensor hardware or high-cost LiDAR systems. The proposed PINMAP framework offers three key advantages. First, PINMAP achieves accurate detection and mapping of transparent obstacles, such as glass, using only low-cost 2-D LiDAR. Second, PINMAP distinguishes between static and temporary obstacles, effectively adapting to dynamic environments. Finally, PINMAP significantly reduces mapping costs by eliminating the need for manual labeling of glass and temporary obstacles. We empirically validate the performance of PINMAP through extensive experiments, including highly dynamic real-world scenarios.