| Recently,indoor agriculture,which cultivates more high-quality agricultural products with a small amount of land,is gradually emerging.Indoor agriculture is a starless environment,such as greenhouse,indoor farms,and orchard tree crowns,where there is no RTK or poor positioning accuracy.Navigation for indoor agriculture has recently become a research hotspot.Due to its robustness and rich information,3D light detection and ranging(Li DAR)has become a popular perceptual sensor and is gradually applied to autonomous navigation of robots.3D Li DAR locates robots based on plants,and plants can affect their growth due to seasonal changes,so on-site testing will be limited by season.In addition,environmental awareness and motion state data cannot be repeated in multiple tests,making algorithm development complex.Based on the above situation,this study proposes a rapid development method for agricultural robot navigation system based on 3D Li DAR for an orchard scene in a starless environment as an experimental environment.Develop autonomous navigation systems in a virtual simulation environment,analyze the characteristics of the field test environment,and design navigation methods suitable for agricultural robots.The main research content of this topic is as follows:(1)Determine the navigation scheme of the robot system and complete the construction of the whole agricultural robot.The environmental characteristics of the test site were analyzed.According to the semi-structured orchard environmental characteristics of the trellis vineyard with slender stems,different shapes and sparse planting,the appropriate navigation sensors,robot system hardware and system software were selected to design and build the overall structure of the agricultural robot.With the center of the agricultural robot chassis as the navigation control point,coordinate unification is carried out according to the installation position of each sensor of the agricultural robot,so that the robot can achieve accurate autonomous positioning between rows when there is a large pose deviation in the process of navigation on bumpy roads.(2)Design robot positioning perception methods and navigation control algorithms.With 3D Li DAR as the main sensing sensor,it senses the point cloud of the surrounding environment.Aiming at the plant and environmental characteristics of the trellis vineyard,it designs an autonomous navigation algorithm for agricultural robot.Through the processing of the 3D point cloud,it obtains the target navigation line of the agricultural robot,and improves the European clustering,least squares and other methods in the autonomous navigation algorithm to improve the navigation accuracy.Finally,it selects an appropriate motion tracking control law,Control the movement of agricultural robot to realize the autonomous navigation of agricultural robot in the trellis vineyard.(3)Virtual simulation environment construction and parameter adjustment.First,according to the environmental characteristics of the trellis vineyard and the basic structure of the agricultural robot,Solid Works is used to reconstruct the trellis vineyard scene,the robot chassis and the 3D Li DAR model;Secondly,use Gazebo to build a unified virtual simulation environment for robots,3D Li DAR,crops;Finally,participate in the verification of the navigation system in a virtual simulation environment to evaluate the reliability of the designed navigation system method.(4)Conduct simulation and field experiments.Using the proposed agricultural robot navigation simulation development method,an autonomous navigation system for agricultural robots has been designed.Simulation,simulation,and field experiments have been conducted to verify the effectiveness of the robot navigation system,and navigation control accuracy evaluations have been given.The results of simulation experiments using GNSS and tracing paths as true values,as well as field experiments in grape sheds,indicate that the maximum absolute errors in the robot’s linear navigation process are 2.72 cm and 3.12 cm,respectively,and the average absolute errors are 1.02 cm and 1.31 cm,respectively,which meet the accuracy requirements for grape planting in greenhouse.Therefore,the proposed agricultural robot navigation simulation development method is effective. |
PDF Full Text Request