Songyi Lu, Quande Yuan, Wenhao Huang, Guoyin Zhang, and Yifan Liu
Laser slam, degradation scenario, graph optimisation, inertial measurement unit (IMU), multiple constraints
In the factory environment, the spatial scale is large and there are many similar scenes. Traditional laser SLAM methods are prone to scene degradation, low positioning accuracy, and large mapping errors. This method is applied in the front-end by incorporating pre- integrated processed inertial measurement unit data and lidar data as constraints into subsequent optimisation. A keyframe is selected at regular intervals, and the best estimated pose is obtained through radar scanning matching. Then the best estimated pose is added to the sub map to estimate the robot’s pose, and cumulative errors are eliminated through loop detection and global optimisation. A comparative experiment will be conducted between the laser SLAM method based on graph optimisation and the method proposed in this paper. The experiment shows that the improved SLAM method based on multiple constraint conditions in this paper can improve the mapping accuracy while reducing positioning errors compared to traditional SLAM methods.
Important Links:
Go Back
