|
With the increasing maturity of UAV (drone) technology, it has been more and more widely used in many fields because of its flexibility. The aim of my project was to improve the mapping and positioning accuracy of in-flight UAVs.
One of the basic tasks of a drone is to determine its location through its own sensors under a given environmental condition. Most mature existing SLAM algorithms assume that the application scene is static and that no moving objects can appear. This assumption limits the application of most traditional or accessible visual SLAM algorithms.
When an unknown environment is large or complex a single drone cannot stably achieve self-positioning and map construction. My project studied a multi-drone VSLAM technology to solve the problem of self-positioning as experienced by the drone. It aimed to solve the defects of the traditional algorithm. A second major benefit of my project was the replacement of the central processing unit by a more efficient model and the introduction of direct UAV to UAV communication. Experiments have showed that the method in this project could solve positioning and mapping problems of drones in dynamic systems and eliminate moving objects with a high success rate to complete the construction of static maps.
Conclusion. As a result of the improvements, the number of point clouds has increased significantly. The drift problem during scanning mapping has been fixed, which makes the algorithm more efficient and accurate.
|