Research On Error State Kalman Filter Fusion Positioning Method In Indoor And Outdoor Mixed Environmen

High-precision localization and navigation are key technologies for intelligent agricultural machinery.Traditional intelligent agricultural machinery commonly uses global satellite positioning(GNSS)based on carrier phase difference technology.In outdoor and unobstructed areas where satellite signals are good,single-sensor positioning can meet the positioning needs of outdoor agricultural production.However,in scenarios such as smart farms or densely vegetated hilly areas,the singlepositioning method cannot meet the requirements due to satellite GNSS signal loss caused by tree and building obstruction or indoor scenarios.To meet the complex indoor and outdoor multi-scene positioning needs of agricultural robots and overcome the limitations of single-sensor positioning,this paper proposes a continuous positioning method based on the combination of GNSS,inertial navigation system(IMU),and multi-line lidar.The work includes the following:(1)Build a multi-sensor data acquisition experimental platform for agricultural robots.First,the overall hardware architecture of the experimental platform is determined.GNSS navigation system,IMU inertial reference system,and lidar are selected for navigation and positioning.The principles and selection criteria of hardware sensors are introduced.Then,by using a nonlinear least squares iterative method,the external parameters of different sensor coordinate systems are calibrated to obtain the corresponding transformation matrices.(2)Research on the fusion algorithm of IMU inertial navigation,lidar,and GNSS to achieve combined indoor and outdoor positioning and navigation.For indoor or obstructed outdoor agricultural environments,this paper uses error-state Kalman filtering algorithm(ESKF)to fuse IMU inertial navigation and lidar for positioning.The accumulated error of IMU inertial navigation is updated through the pose information of the prior point cloud map observed by lidar to obtain the optimal posterior pose estimation,thereby improving the positioning accuracy and system robustness.For open outdoor agricultural environments,this paper uses GNSS and IMU for fusion positioning,and the integration data of IMU can effectively correct the data deviation of GNSS and improve the positioning accuracy.(3)Research on the lidar point cloud registration algorithm and propose an improved FPFH-ICP point cloud registration strategy.In the point cloud map positioning module solution for indoor environments,global map matching needs to be performed through the feature of local point cloud frames to obtain the global pose information.In addition,this method is also used for the global initialization of prior map positioning.(4)At the alternating point between the indoor and outdoor positioning modules,the positioning strategy needs to be switched according to the scene changes.On this basis,this paper proposes a method to select the module with higher precision by comparing the posterior weight values.The weighted sum of the corresponding diagonal elements in the posterior covariance with the location information that meets the conditions is used to obtain two weight values for indoor and outdoor,and the module with the larger weight value is selected as the positioning module at the alternating point.