Research On Greenhouse Mobile Robot Navigation Based On Multi-sensor Fusion

With the continuous expansion of the number and scale of greenhouses,the contradiction of insufficient labor force and population is becoming increasingly prominent.How to achieve automated operation of robots in greenhouses has become one of the current research hotspots.Due to the fixed navigation path of commonly used greenhouse mobile robots,high sensor costs,and poor satellite signal positioning within the greenhouse,achieving a greenhouse mobile robot navigation system is a prerequisite for completing greenhouse robot automation operations.This article studies the relevant SLAM(Simultaneous Localization And Mapping)technology and path planning algorithms,builds a greenhouse mobile robot platform,and designs a greenhouse mobile robot mapping and navigation system based on multi-sensor fusion.The main job responsibilities are as follows:Firstly,based on the characteristics of the greenhouse environment for vegetable crops,the overall architecture of the greenhouse mobile robot navigation system is designed,and the system hardware is analyzed and selected.Then,according to the greenhouse operation requirements,the chassis structure of the greenhouse mobile robot is defined,and the two wheel differential mobile chassis is selected to establish the differential kinematics model and sensor observation model.Finally,by comparing the characteristics of commonly used environmental map models in navigation,a grid map is selected as the greenhouse environment map.Next,the mainstream laser SLAM algorithm was studied,and greenhouse environment simulation mapping experiments were conducted on the filtered Gmapping and graph optimized Cartographer algorithms.The mapping accuracy was compared,and the Cartographer algorithm was ultimately selected.Due to the fact that the Cartographer algorithm only uses a single odometer information for pose estimation,there may be inaccurate pose estimation issues.Therefore,the error performance of Extended Kalman Filter(EKF)and Unscented Kalman Filter(UKF)in pose estimation was compared,and UKF was selected as the fusion pose estimation algorithm.Integrate IMU information into the laser odometer model of Cartographer,and use UKF for information fusion to achieve prediction of the pose of mobile robots.Validation was conducted on the MIT Stata Center dataset and simulated greenhouse environments,and the results showed that the multi-sensor fusion mapping results were better after adding IMU information.Then,design and improve a path planning algorithm that integrates A* and DWA.To address the issues of low security and obstacle avoidance in traditional A* paths,dynamic weighting and new search strategies are adopted to improve path search efficiency and solve the problem of unsafe planned paths.In response to the problem of local path planning DWA algorithm falling into local optimal solutions,the evaluation function of the DWA algorithm is improved.The improved A* is fused with DWA,and reference points in the global path of the improved A* planning are selected for segmented local path planning to complete robot navigation obstacle avoidance.The effectiveness of the algorithm is verified in MATLAB and ROS.Finally,a greenhouse mobile robot platform with sensors such as wheel odometers,lidar,and IMU was built based on the Robot Operating System(ROS)to complete greenhouse mapping and navigation tasks.Firstly,analyze the reasons for the information error of the robot odometer and complete the calibration.Subsequently,due to the problem of motion distortion when using low frame rate lidar to collect data,a wheel odometer was used to assist in lidar motion distortion correction.Afterwards,due to the high computational power requirements for the operation of Cartographer,distributed improvements were made to it.Then,navigation mapping experiments were conducted in the greenhouse scene,and the results showed that the multi-sensor fusion mapping with IMU information can meet the needs of navigation;The integrated navigation algorithm can achieve obstacle avoidance for robot navigation.Finally,SLAM navigation experiments were conducted in actual greenhouse scenarios,and the position deviation and heading deviation were higher than the simulated environment.The position deviation was less than 15 cm,and the heading deviation was less than 14°,meeting the requirements for greenhouse operations.Therefore,it has been verified that the multi-sensor fusion mapping navigation scheme is feasible.