Research On Map Construction And Path Planning Of Orchard Mobile Robot

At present,China’s forest fruit production has become a new highlight and important industry for the economic development of most forest fruit industrial areas and farmers to increase their income and get rich.However,fruit production in the country is facing industrial difficulties such as low mechanization level,mismatch of agricultural machinery and agronomy,and high labor cost.In order to improve the automation level of orchard industry,introducing autonomous navigation technology into orchard operation tasks becomes an effective way to solve such industrial problems.In this paper,we focus on the map building and path planning technologies in the autonomous navigation of orchard mobile robots,and investigate the robot operation system(ROS),2D Li DAR-based Simultaneous Localization and Map building.The ROS is used to build an orchard environment model and a mobile robot model equipped with Li DAR,and to conduct map building simulation experiments.We propose metrics to evaluate the map building accuracy of the algorithm.For the path planning of orchard environment,we propose a hybrid path planning method with additional navigation points and conduct simulation experiments to verify its reliability.Experiments on map building and navigation are conducted using a small mobile robot in kind.The main work and conclusions of this paper are as follows:(1)For the problem of map construction in standardized orchards,we study the Gmapping,Cartographer algorithm based on 2D Li DAR.In ROS,we write URDF file to create orchard environment model,mobile robot model and Li DAR model,load the model in 3D physical simulation software Gazebo,visualize the information using Rviz software,and call SLAM algorithm function package for orchard environment map construction experiment.The map constructed by SLAM algorithm and the original map were processed,and three indexes,namely mean square error,mean error of trunk center of mass and trunk intersection ratio,were set to compare and evaluate the accuracy of the map constructed by SLAM algorithm.The simulation experiments show that in the structured standard orchard environment,the average size deviation of the map constructed by Gmapping is 0.425 and that of the map constructed by Cartographer is0.0725.The accuracy of the map constructed by Cartographer is higher than that of Gmapping.(2)To address the problem that the orchard mobile robot has more repeated turns in the orchard arranged in tree rows and the path search efficiency is low,the orchard corridor map construction and navigation point generation method is proposed,the tree rows are expanded to convert the discrete fruit trees into a continuous narrow corridor model,and according to the characteristics of this environment,the intermediate navigation points are inserted at the turns to transform the path planning problem from the starting point to the end point into the path planning between the navigation points problem.Dijkstra and A* algorithms based on graph structure are studied,and A* algorithm is selected as the global path planning algorithm after comparing the algorithm performance.The local path planning algorithm uses the dynamic window method.Hybrid path planning simulation experiments of obstacle avoidance experiments and increasing navigation points are conducted in ROS.The experiments show that the hybrid path planning method with increased navigation points can plan a passable and safe path with large turning radius,less number of turns and centered driving,which meets the requirements of orchard operation.(3)In order to simulate the actual orchard environment,a small mobile robot was used for simulating orchard mapping and navigation experiments.The robot is composed of an upper computer and a lower computer.The upper computer includes a two-dimensional Li DAR,Raspberry Pi4 B,and a WIFI module.The lower computer is composed of STM32,a motor drive module,and an incremental encoder.It communicates with a laptop through WIFI using an SSH program.2D Li DAR collects tree trunk information,incremental encoder accumulates mileage information,uses Gmapping to construct a map,and generates a simulated orchard environment map.And using this map as a prior map,load it into ros navigation,run the Python program for setting navigation points,and use a hybrid path planning method that adds navigation points to generate planned paths between navigation points.The average navigation deviation value is4.5975 cm,and the mobile robot can drive autonomously along the navigation path.