A Laser SLAM Algorithm for Fusion of Visual Line Features

The algorithm proposed in this manuscript improves the positioning accuracy, closed-loop capability, and scene construction effect of the 2D laser SLAM algorithm by introducing environmental line features. Comparative analysis with different SLAM algorithms on the OpenLORIS dataset and recorded data packets in real environments shows that the positioning accuracy and closed-loop ability of this algorithm have been significantly improved in various environments, and a more informative environmental structure map has been obtained. The OpenLORIS dataset sequence folder contains the following: OpenLORIS dataset(https://lifelong-robotic-vision.github.io/dataset/scene)1. Groundtruth trajectory truth value 2. Hector algorithm pose TUM data 3. ORB-SLAM3 ORB-SLAM3 algorithm pose TUM data 4. PL-VINS PL-VINS algorithm pose TUM data 5. PL-VIO PL-VIO algorithm pose TUM data 6. Cartographer Cartographer algorithm pose TUM data 7. Output TUM data for the pose proposed by OurMethod The mybag package folder contains the following: 1. True value of ArUco trajectory 2. Hector algorithm pose TUM data 3. Cartographer Cartographer algorithm pose TUM data 4. Output TUM data for the pose proposed by OurMethod Script for evaluation (using evo tool for trajectory comparison) 1. Result.sh comparison from evo_ Ape and evo_ Tools for one or more result files generated by RPE 2. Traj.sh Compare Trajectories 3. APE.sh absolute pose error comparison