| With the rapid development of science and technology,especially MEMS technology,the cost of rotorcraft UAV is getting lower and lower,and its application in civil and industrial aspects is also increasing gradually.Therefore,the full autonomy and intelligence of UAVs become more important.Real-time accurate positioning in unknown environment and subsequent navigation technology are one of the most important links.Aiming at the outdoor environment,this paper uses the multi-line lidar as the main perception mode and the inertial navigation module as the assistance to study the three-dimensional real-time positioning and environmental map construction of the rotary-wing UAV.The main research contents include the construction of the unmanned aerial vehicle platform carrying lidar and IMU,the pre-processing of lidar original point cloud,the pre-integration of IMU,and the back-end optimization method based on factor graph.Finally,the data collected in the real environment are used to compare and verify different algorithms.For multi-line Li DAR point cloud data were collected with moving distortion and interference problems,analyzed the movement distortion as well as the causes of noise points,and by using innovative marketing information on the laser point cloud movement distortion correction,through a variety of filter way,after several noise filtering,this method can effectively reduce the effects of noise on feature extraction.In this paper,the feature-based point cloud matching algorithm,laser odometer and IMU pre-integral inertial odometer are realized,and the ground height calculation based on Li DAR is realized.In this paper,a map construction method based on the sliding window method is implemented.The map is decomposed into the key frame point cloud and the corresponding key frame pose is sav ed.Due to the continuous error accumulative in SLAM algorithm,this paper uses Euclidean distance between key frames to detect loop,and uses ICP matching constraints to calculate the relative pose between key frames.Then,the incremental smoothing optimization method based on factor graph and Bayesian tree is used to obtain the current UAV pose,which effectively reduces the positioning error.In this paper,a six-rotor UAV platform equipped with Li DAR and high-precision IMU was used to collect multiple sets of data,and RTK was used as the position benchmark to evaluate the odometer obtained by this paper and a variety of other algorithms,including absolute position error,relative position error and mapping effect.The results show that the proposed algorithm is superior to the current mature algorithms,and can meet the practical application.In general,according to the actual flight environment of the UAV,this paper adopts the method of IMU and height data,and uses factor graph and incremental smoothing optimization to improve the problems of high insensitivity and large error accumulation of the previous laser SLAM system,which not only improves the accuracy,but also optimizes the mapping effect.Finally,a number of experiments are carried out to verify that the proposed algorithm can meet the localization requirements in the real environment. |
