Autonomous Navigation Of Orchard Track Mobile Platform Based On Multi-sensor Fusion

China’s forestry and fruit industry is developing rapidly.At present,most of the orchard operations are semi human or pure human operations,and lack of intelligent mobile platform.When the intelligent mobile platform carries out inter row operation,the position accuracy of the vehicle is very important.Accurate positioning information can not only provide the environment and its own position information for the mobile platform,but also determine the navigation route and obtain the navigation parameters on this basis.The traditional GPS positioning system relying on high-precision and high-frequency output is expensive,and once it fails,it will lead to the failure of the whole platform.To solve this problem,this paper studies and designs a multi-sensor fusion of autonomous navigation of agricultural tracked mobile platform.Taking the agricultural crawler mobile platform as the test platform,several key technologies in the navigation system are studied(1)The existing crawler mobile platform is transformed,and the self-developed automatic control system is added to the mobile platform.Through the controller to control the electromagnetic directional valve to control the expansion and contraction of the hydraulic cylinder,the vehicle motion control is realized.(2)The coordinates of fruit trees are obtained by lidar and binocular vision and converted to the base coordinate system of crawler mobile platform,which is used to fit the path.Combined with lateral deviation and heading deviation,the linear navigation between lines is completed.Based on the lidar detection of fruit trees,the obtained data were processed to obtain the trunk coordinate information;Through laser marking,combined with binocular camera,the image was collected and processed to obtain the tree trunk coordinates;The two coordinates are transformed into the base coordinate system of the mobile platform,and the least square method is used for path fitting.The course deviation is obtained by heading and attitude sensor,and the lateral deviation is obtained by lidar.Experimental results show that compared with single sensor navigation,this method can improve the navigation accuracy(3)Compare with the headland coordinates,judge whether it is in the headland,confirm that after entering the headland,use the way of straight curve to transform the headland turning arc path into a straight path.The GPS coordinates are compared with the coordinates of the starting point of the earth’s head to determine whether the earth’s head is reached.At this time,GPS is used to turn the mobile platform.The GPS coordinates are compared with the coordinates of the end point of the ground to determine whether to end the turn.The strategy of turning is to divide the semicircle model into several straight lines and turn turning driving into straight driving.The least square method is used to fit the GPS data to complete the path recognition.The test shows that the actual driving path is basically the same as the planned path when turning on the ground.(4)With vs as the development platform,software design and debugging are completed.Through the fusion of lidar,GPS and binocular vision information,the navigation path detection is completed.According to the control strategy of straight line and ground turn,the program development of the whole navigation system is completed,and the line navigation and ground turn of tracked mobile platform are realized.