Research On Autonomous Operation Mobile Platform System Based On Facility Environment

Facility agriculture provides an important way forward to realize agricultural intelligence,which can improve land utilization and quality of agricultural products and enhance comprehensive agricultural production capacity.Facility agriculture is mainly based on smallscale operation equipment,which has such challenges as high operation intensity and harsh environment,so it is necessary to design an autonomous mobile platform for the facility environment.The existing navigation technology cannot be applied to the closed environment of facilities,and other positioning and navigation technologies need to be sought.Simultaneous Localization and Mapping(SLAM)is an existing technology that relies on sensors for autonomous localization and environment sensing,among which laser SLAM has the characteristics of robustness and wide applicability,however,the facility environment crop is complex and the amount of laser point cloud data features is sparse,so it cannot be directly navigated.By exploring the deficiencies in single-sensor environment construction and studying the laser 3D tightly coupled map building method that fuses visual information,this topic constructs a localization and map building method with high accuracy and good robustness to provide a theoretical basis for facility environment navigation on mobile platforms.The specific research contents are as follows:(1)Structural design of the autonomous mobile platform according to the environmental conditions of the facility,including movement,steering methods and suspension forms.Build an upland gap autonomous mobile platform hardware and software system to package and configure different control modules.At the same time,the kinematic analysis of the four-wheel differential model of the autonomous mobile platform is carried out to solve the motion control of the platform in order to realize the autonomous navigation operation in the facility environment.(2)To address the shortcomings of single sensor in localization building,the visual data and LiDAR sensor data are fused,the bidirectional RANSAC method is proposed to eliminate the visual feature mis-matching point pairs,the 3D-2D and 3D-3D error models are jointly constructed for iterative solving during the positional solving process,and the loopback framework is added to eliminate the drift error,and the simulation dataset is tested to verify the method has excellent accuracy and robustness in constructing multi-closed-loop maps for large scenes;then the key localization techniques and path planning techniques in autonomous navigation are studied and analyzed,and the A*algorithm and dynamic window method(DWA)are selected for path planning by combining AMCL and odometer localization methods,and through simulation tests,it is shown that the best path can be successfully planned safely and quickly in the environment,while being able to real-time obstacle avoidance.(3)To verify the feasibility of the method of this topic,the system level of the mobile platform is built and configured with relevant parameters,and the platform is used to build a map test in the facility scene to verify its loopback detection performance and environmental adaptation capability in the construction of a large scene environment;subsequently,the built highland gap autonomous mobile platform is used in the facility environment for path planning tasks at different locations,straight-line driving and steering driving at different speeds Position error and heading angle error analysis,autonomous mobile performance study under different expansion radius.The test results show that the mobile platform can complete the tasks of environment construction,positioning and autonomous navigation,and can meet the requirements of the mobile platform for facility environment operation.