KP-Cartographer: A Lightweight SLAM Approach Based on Cartographer

Cartographer is a simultaneous localization and mapping (SLAM) approach proposed by Google, capable of creating maps and localization from point cloud data obtained from LiDAR. In order to perform localization and mapping on mobile devices with lower computational power, suitable for scenarios such as firefighting and rescue, we designed a laser point cloud feature extraction algorithm and a personnel localization algorithm. We integrated them into the Cartographer method, proposing a lightweight SLAM method named KP-Cartographer. To validate the effectiveness of our method, we constructed five laser point cloud datasets with different characteristics and conducted multiple comparative experiments. Qualitative experimental results demonstrate that the KP-Cartographer method proposed in this paper surpasses A-LOAM and LeGO-LOAM in map completeness and clarity. Compared to Cartographer, it effectively identifies small obstacles and accurately captures differences between glass and window frames. Quantitative experimental results indicate that, compared to Cartographer, 1) The KP-Cartographer method proposed in this paper reduces the volume of laser point cloud data to within 6% of the original data, with a maximum CPU utilization reduction of 21.96%. 2) The maximum Absolute Trajectory Error (ATE) of the KP-Cartographer method proposed in this paper is 0.0308, and the maximum Relative Pose Error (RPE) is 0.0068. This achieves lightweight processing with no significant difference in accuracy.