| In order to improve the mechanization and intelligence of Chinese cabbage harvesting and realize the automatic navigation of field harvesting,an automatic navigation system suitable for unstructured field operation was studied based on 3D LiDAR,and its tracking accuracy and obstacle avoidance performance were tested in the field.Mainly completed the following work:(1)Construction of navigation system platform and related theoretical basis.This paper analyzes the shortcomings of mechanization operation of Chinese cabbage at home and abroad,summarizes the research status of automatic navigation of agricultural machinery.The navigation system is divided into decision control module,navigation and obstacle avoidance module and motion control module by using remote sensing and automatic navigation on the test platform.According to the unstructured field environment and navigation requirements,key components of the test platform were chosen and the test platform was established.The software system of the test platform based on ROS framework was completed.3D LiDAR was constructed as the partial frame of reference system of the test platform by employing the coordinate system transformation between sensors.(2)Research on ridge path planning based on 3D LiDAR.3D LiDAR and IMU were used for information fusion to complete the ridge row point cloud extraction.The data of ridge row point cloud were down sampled and statistically filtered by voxel filtering,and 67.8%of outliers and outliers were eliminated.Then,according to the effect of eliminating ground point cloud by GPF,CSF and PMF algorithms,the GPF algorithm with 94.9%extraction rate of ground point cloud was selected.SLR algorithm was used for cluster segmentation of ridge row point cloud,to strengthen the distribution characteristics of ridge row point cloud,and to set ROI in the region of interest.In order to obtain more accurate row path,RANSAC,LSM and Hough algorithms were compared to fit the row line,and the feasibility and real-time performance of fitting were comprehensively considered.RANSAC was selected as the fitting method of navigation path.(3)Research on ridge path tracking based on PID.According to the crawler differential motion model,the motion mode of the test platform is analyzed and the kinematic model is built.The PID control is used to realize the trajectory tracing of the ridge line,and the relationship between the left and right crawler wheel and the speed is determined by the deviation of the navigation Angle,so that the communication protocol instruction is sent to control the movement control of the test platform.In order to enhance the security performance of the test platform,the TEB local obstacle avoidance algorithm is used to detect obstacles to ensure that the test platform timely detects and stops moving before a certain safe distance.(4)Field experiment results and analysis of navigation system.The feasibility and safety of the navigation system are verified by ridge tracking accuracy test and obstacle detection test.The experiment consequence indicate that when the test platform is traced at the velocity of 0.2 m/s,0.4 m/s and 0.6 m/s,the test results are the best at 0.2 m/s,the average value of transverse and longitudinal deviation is less than 6 cm,the maximum value is less than 10 cm,the average value of navigation Angle deviation is less than 4°,the maximum value is less than 9°,in addition,The average response time and distance to the obstacle of the test platform are less than 181 ms and 1.3 m,and the linear model between them is obtained.The results showed that the automatic navigation system could satisfy the automatic navigation of Chinese cabbage in the field and had certain safety. |
PDF Full Text Request