Research On The Full-field Automatic Navigation System Of Agricultural Machinery Based On Multi-source Information Fusion

As China’s urbanization process advances rapidly,the area of arable land is increasingly shrinking,and a large number of rural young and middle-aged labor force is flocking to cities,resulting in a serious shortage of rural farming labor and an increasingly prominent situation of few people or no one farming.Therefore,the transformation of traditional agricultural production mode to smart agriculture and unmanned farms is imminent,and intelligent agricultural machinery is an important equipment support for the construction of unmanned farms.The automatic navigation system of agricultural machinery is an important part of intelligent agricultural machinery equipment.Improving the reliability and intelligence level of automatic navigation of agricultural machinery is of great significance for promoting the national rural revitalization plan and entering the strong agricultural country.This paper studies several aspects of agricultural machinery automatic navigation technology,such as positioning perception,multi-source information fusion,obstacle information detection and path tracking,and integrates the development of automatic driving system to verify the relevant algorithms.The specific contents are as follows:(1)Heading information fusion based on Kalman filtering.In view of the problems that the heading output of dual-antenna RTK positioning and orientation system is easily disturbed by signal interference,antenna installation error and limited baseline length,a method based on Kalman filtering is proposed to realize RTK-GNSS/AHRS heading data fusion by estimating AHRS heading error.The test results show that this method can effectively improve the heading output stability of positioning and orientation system under satellite signal occlusion.(2)Obstacle detection using low-cost radar.In view of the problem that agricultural machinery may encounter obstacles in unstructured field environment,a radar host computer software with obstacle detection function is developed based on DJI Livox Mid-40 laser radar,and the automatic parking and obstacle avoidance function of agricultural machinery is realized through the information interaction between radar host computer software and navigation host computer software.The test results show that this method can effectively monitor the real-time position of obstacles and stop and avoid obstacles within the set detection distance.(3)Development and testing of low-cost positioning and orientation system.In view of the expensive price of RTK positioning and orientation system in the market and the lack of integrated base station data receiving module,an RTK-GNSS positioning and orientation system that can output high-precision positioning information in real time is developed based on embedded platform and Free RTOS real-time operating system,integrating UM482 positioning and orientation module and DU1018 D data radio module.In view of the problems of low satellite positioning data update frequency and inability to obtain real-time attitude information,a GNSS-IMU combined positioning system integrating UM482 positioning and orientation module and IMU102 N inertial measurement unit is developed,which can realize the synchronous acquisition and integrated transmission of GNSS and IMU data.The test results show that both positioning and orientation systems can output the expected information and work stably for a long time.(4)Path tracking based on improved Stanley model.In view of the poor adaptability and low curve tracking accuracy of the commonly used Stanley model for agricultural machinery path tracking,a PSO-FSM path tracking algorithm is proposed to adaptively adjust the gain coefficient of the Stanley model using fuzzy algorithm and further optimize the expected steering angle by particle swarm optimization.The mobile car field test results show that when the initial error is 4 m,the car’s on-line distance does not exceed 5 m,and the maximum error of the whole field path tracking is 3 cm.The field autonomous operation test of John Deere C230 combine harvester shows that the maximum error of the whole field path tracking is 0.63 m,which meets the demand of path tracking accuracy for unmanned operation of combine harvesters with large working width.