UAV Autonomous Navigation Technology Based On Fusion Of LIDAR And IMU In Double-blind Environment

With the rapid development of micro UAVs in recent years,they have been widely applied to real life due to their flexibility,mobility and low cost.Generally,UAVs' navigation states are usually estimated by fusing IMU and GPS,and vision sensors,and LIDARs,when flying autonomously.As the expansion of UAVs' application area,applications in “double-blind” environments where there is no available GPS signal and light gradually increase.Navigation methods based on GPS or vision will both fail in “double-blind” environments,but LIDARs,which operate independent of external illumination,can obtain accurate ranging data in dark environment.Thereby,LIDARs favor UAV autonomous navigation more.Because of the size,weight and power limitation,2D LIDAR is more suitable for small UAVs.Main problems when using 2D LIDARs to localize UAV with 3D motion are: 1)In the environment of abrupt structural change in altitude direction,the scanned 2D environment structure may vary with UAV's height,which contradicts the assumption that perceived environment is not dynamic in traditional algorithms.This causes large error in pose estimation.2)In environments like “single plane”,because the lack of reference in one direction,previous methods using LIDAR can't correctly estimate drones' motion in the direction lacking reference.In order to solve problems above,the autonomous navigation technology for UAVs based on LIDAR and IMU in double-blind environment is studied in this paper.Firstly,the error characteristics of pose estimation for classic algorithms using LIDAR in the environment exists dramatic structural change in altitude direction and in the environment there is only “single plane” are analyzed.Then,taking traditional methods' error characteristics into account,an algorithm based on fusion of LIDAR and IMU is proposed to solve these problems in the complex environment.With LIDAR scan data,identify whether current environment is “single plane” environment first,As soon as current environment varity transforms,estimator adopts the method corresponding to current environment to ensure precision and reliability of navigation during the whole flight in complex environment.Aiming at the problem that large positioning error of previous SLAM methods using 2D LIDAR in the environment where there is structural step changes in the altitude direction,an IMU aided Robust-SLAM algorithm is proposed to estimate UAV's states,and that a reliable 3D SLAM can be accomplished with 2D LIDAR.Accurate environment step change detection is introduced into SLAM,which combines IMU and point cloud information,and pose estimation can be acquired with the states predicted from IMU assisting LIDAR.A 3D environmental map is constructed with the pose from fusion.In order to deal with large positioning error of the estimation using LIDAR in “single plane” environment on account of the lack of parallel direction reference feature,a dynamic model/LIDAR/IMU integrated algorithm is proposed to estimate UAV state.Firstly,line characteristic information of “single plane” is extracted from LIDAR point cloud to observe the distance and yaw angle of UAV relative to “single plane”,and the velocity of UAV is observed by dynamic model algorithm.Lastly,accurate navigation states of UAV are attained from the fusion of above observation information and IMU.In this paper,an autonomous navigation algorithm based on 2D LIDAR verification platform is constructed on DJI M100 rotorcraft,and proposed methods are realized,what's more,a control approach respect to the navigation algorithm is designed.Thus,a closed-loop for navigation,guidance and control is formed.The accuracy of the proposed navigation method is proved by indoor flight tests with the constructed UAV platform.In addition,some flght experiments in a large storage environment have been carried out,which demonstrates good environmental adaptability of the UAV system and navigation algorithm.