Research On Mapping And Navigation Of Environment Inspection Robot In Animal Husbandry Farm

Modern technology leads to the constant reduction of the cost of lidar and vision sensor,and certain breakthroughs have been made in technology.The research on autonomous navigation robot has become a hot research field.In practical applications,autonomous navigation robots are no longer only experimental products in the laboratory environment,but gradually begin to step out of the laboratory to carry out industrial production and animal husbandry and other industries.Autonomous navigation is an important link to realize unmanned industry.In order to achieve autonomous navigation,the robot needs to know its position according to the sensor,and the required position requires a set of relatively accurate maps.The two technologies are generally known as simultaneous positioning and map construction(SLAM).In this paper,aiming at the environment of animal husbandry plants,the traditional wheeled autonomous mobile robot was improved with ROS as the framework,and the construction of tracked mobile robot system and relevant SLAM technology were studied.Firstly,the hardware system required by the system was built,and the mechanical structure of the crawler movement platform for the inspection robot in the animal husbandry plant was designed.The environment map model of SLAM system,the basic principle of constructing raster map and the sensor model are described.Based on the analysis and research of the filter based SLAM algorithm,a variety of two-dimensional and three-dimensional SLAM algorithms were compared and tested,which laid a foundation for the construction of multi-sensor fusion technology.The VINS_Mono algorithm was improved under the ROS framework.The improved VINS_Mono algorithm and the graph optimization algorithm could have strong robustness in the animal husbandry environment,so as to achieve the purpose of application.In the optimization of the algorithm,the environment to be built by analyzing the map.The improved map optimization algorithm and the improved visual VINS_Mono algorithm are used to extract the feature points in the map construction.The traditional feature extraction method is depth information,which can only extract the points and boundary lines of the environment.As a result,multiplesemantic information on the map is lost,resulting in inaccurate sparse maps.To improve the above problems,the extraction of the loopback detection features were added to the algorithm model,put forward the vision algorithm and the traditional single laser radar hybrid parallel computing of the extraction algorithm,the improved algorithm to build the simulation experiment of breeding environment map model compared with the actual environment,found by comparing the experimental results of model and the actual experimental environment closer to achieve the goal of hybrid parallel algorithm.At the same time,it solves the problem that the blind area in the simulated breeding environment cannot be detected.The basic theory of SLAM based on filtering method and graph-based optimization method is mainly used.The basic principles of SLAM of the two methods are introduced and simulated.Gmapping,Hector SLAM in filtering algorithm and Cartographer in cartographic optimization algorithm are introduced and compared experimentally.At the same time,ORB-SLAM algorithm model and VINS_Mono algorithm model were used in the experiment of visual SLAM algorithm.