Study On Localization Of Legged Robot Using Multi-sensor Fusion

Autonomous localization is the core function of legged robot.Compared with the conventional wheeled robot,legged robot is driven by leg joints to achieve robot motion.Its special form of motion makes it challenging to obtain localization information,including the calculation of legged odometer,pose estimation in the case of robot motion shaking,accurate localization in large-scale outdoor scene,etc.In order to realize the autonomous localization of legged robot,this paper design a multi-sensor fusion localization method based on the existing robot localization methods,mainly studies how to use the robot body sensor combined with external observation information to achieve robot localization,and carry out the related experiments.Works of this thesis are as follow:(1)Aiming at the problem of calculating the odometry of leg foot robot,a leg foot odometer based on robot kinematics model is designed.Firstly,the structure of the quadruped robot is analyzed and the kinematics model of the robot is established.Then the extended Kalman filter is used to fuse the kinematics information and IMU measurement information to realize the legged odometer localization algorithm,and the error sources of the legged odometer based on kinematics are analyzed.Finally,the experiment verifies that the leg foot odometer estimation based on kinematics and IMU can roughly get the robot mileage estimation Accounting information.(2)The odometry estimation algorithm of IMU based on lidar point cloud fusion is realized,which provides laser mileage information for subsequent sensor fusion.Firstly,the lidar point cloud is preprocessed,and the motion distortion of the point cloud is removed by fusing IMU;secondly,based on the iterative closest point algorithm,the pose matching algorithm based on the distance between point line and point plane is adopted;finally,the pose is further optimized by point cloud and local map matching.(3)Aiming at the problem of robot multi-sensor fusion localization,a set of sensor fusion framework based on graph optimization is designed,which realizes the effective use of sensor information and the improvement of robot localization accuracy.Firstly,the robot localization problem is modeled,and the sensor fusion localization framework is designed;secondly,the sensor information is synchronized in time and space;secondly,the position and pose optimization of the robot odometer is preliminarily realized based on the legged odometer fusion laser odometer information;finally,the GNSS positioning and closed-loop detection are fused based on the legged-lidar odometer to achieve more accurate localization of the robot.Compared with simple odometer localization,the average localization error of legged-lidar odometer fused with laser odometer information is 4.94 meters,which improves the localization accuracy by 53.7%;the average localization error of multi-sensor fusion localization is 1.06 meters,which improves the localization accuracy by more than 78.5%.The experimental results show that the multi-sensor fusion localization algorithm can greatly improve the localization accuracy of legged robot.