Applied Research Of NDT Algorithm In Small Four Rotor UAV Location

In recent years,the four rotor UAV has been widely applied as a commonly used multi rotor UAV.The position estimation of four rotor UAV is always a research hotspot and difficulty in the field of UAV.It is also the basis and premise for completing various complex tasks.The appropriate location algorithm is the key to UAV navigation.The commonly used location algorithms are Iterative Closest Point(ICP),normal transformation algorithm(Normal Distributions Transform,NDT)and so on.Because the iterative closest point algorithm needs to extract feature points and match feature points,so the algorithm is inefficient and slow.The normal distribution transformation algorithm is a match determined by the whole factors.It avoids complex problems of point to point in registration process.Compared with iterative closest point algorithm,the algorithm is efficient and fast.This topic mainly studies the location of NDT algorithm in small four rotor UAV.In this project,we first use the UTM-30 LX 2-D laser scanning rangefinder carried by the four rotor UAV to get the distance and angle information of the discrete points in the environment.Then it is converted to two-dimensional data points in the world coordinate system,and point clouds are obtained.Secondly,we preprocess the obtained point cloud,compare the data structure of Kd-tree and Octree,use Kd-tree data structure to organize the data organization of point cloud,compare Euclidean distance method and K neighborhood search method,choose Euclidean neighborhood search method to complete the nearest neighbor search,and then compare based on distribution,depth,clustering based,based on Distance and density based outlier discrimination method,the distribution based Statistical Outlier Removal outlier filtering algorithm is used to filter out the outliers,and the pre processing work of point cloud is completed.Finally,the pre processed laser scanning data is matched by the NDT algorithm,and the 2D discrete data set by the laser scanning is converted into a piecewise continuous and differentiable probability distribution in the two-dimensional plane.We match the probability distribution of target point cloud and source point cloud,calculate the sub point of matching points,and use Newton algorithm to optimize,and get the best transformation parameters.The positioning test platform of UAV is set up.The hardware platform uses Pelican four rotor UAV equipped with UTM-30 LX 2-D laser sensor.The softwareplatform is Linux Ubuntu 64 bit operation system under PC.Install Linux Ubuntu 64 bit operation system on PC,install Qt software and configure PCL library,then locate test.It is verified that the NDT algorithm can accurately estimate the motion of UAV and achieve the location of NDT algorithm in small four rotor UAV.