研究目的
To compare the performance of three ROS-based 2D SLAM algorithms (Google Cartographer, Gmapping, and Hector SLAM) in terms of map accuracy against a precise ground truth.
研究成果
Google Cartographer generally produces maps with the smallest error compared to the ground truth across various movement conditions. Gmapping performs reasonably well, especially without loop closure, while Hector SLAM is less accurate due to its reliance solely on lidar data and lack of explicit loop closure. The method is suitable for extending to 3D map comparisons.
研究不足
The study is limited to 2D SLAM in a static indoor environment; it does not address 3D SLAM or dynamic environments. The comparison uses only the ADNN metric, which may not capture all aspects of map quality.
1：Experimental Design and Method Selection:
The study involved comparing three SLAM algorithms using the ADNN metric. Experiments were conducted in a static indoor environment with the robot moving under different conditions (slow/smooth, fast/smooth, fast/sharp, no loop closure).
2：Sample Selection and Data Sources:
Data was collected from a mobile robot equipped with a 2D lidar in a classroom environment at Innopolis University.
3：List of Experimental Equipment and Materials:
Mobile robot 'Plato' with Jetson TX1 computer, Hokuyo URG-04LX-UG01 2D lidar, FARO Vantage Laser Tracker, Metrolog X4 software, ROS Kinetic, Ubuntu 16.
4：Experimental Procedures and Operational Workflow:
04.
4. Experimental Procedures and Operational Workflow: The robot was teleoperated using a joystick to record ROS bag files. These files were processed offline by each SLAM algorithm to generate occupancy grid maps. The ground truth map was constructed using the FARO laser tracker and Metrolog X4 software by measuring 3D points and extracting a 2D slice at lidar height.
5：Data Analysis Methods:
Maps were aligned using the ICP algorithm, and the ADNN metric was calculated to compare the SLAM-generated maps with the ground truth.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
