Design Optimization Of SLAM Parameters For AGV Vehicle Based On Steering Modes

The AGV(Automated Guided Vehicle)is a vehicle that can move automatically at a specified routine and is currently used in industrial manufacturing,logistics and home service applications.With the development of productivity and the improvement of living standard,the application prospect of AGV is becoming more and more promising,which requires the further development of AGV in the three core technologies of environment awareness,path planning and motion control.Environment awareness technology is a prerequisite for the realization of AGV autonomous navigation capability.It can help the AGV to detect external environmental information and determine its relative position.Path planning is the key to AGV autonomous navigation,which requires the AGV to obtain the location information of itself and the navigation target on the map.Simultaneous Localization and Mapping(Simultaneous Localization and Mapping)can be used to address context-aware issues and assist AGV in locating or Mapping capabilities,thus laying the foundation for further path planning.Before controlling the motion,we need to understand what objects the AGV controls.For example,under different steering forms,the AGV controls different objects.There are various modes of AGV steering in the current market.Two-wheel differential steering,four-wheel differential steering and Ackermann steering are the three most common forms of steering.Among them,the widely used laser SLAM algorithm is designed on the basis of two-wheel differential steering,and many parameters of SLAM are generally based on this steering mode,based on experience or for convenience,and there is room for optimization of its parameters.At the same time,there is room for improvement in the back-end optimization algorithm of SLAM.Therefore,this paper first discusses the advantages and disadvantages of four-wheel differential steering and Ackermann steering in the path control,and then,when the steering mode is four-wheel differential steering,the parameters and back-end of SLAM are optimized,so as to reduce the positioning error of SLAM and improve the mapping effect.In this paper,the kinematics modeling of four-wheel differential steering and Ackermann steering was carried out first,and the curvature formulas of four-wheel differential steering and Ackermann steering AGV were obtained.Then,the basic principle of SLAM is analyzed.Through an example,the positioning accuracy of particle filter and extended kalman filter in filter-based SLAM is compared.The result shows that the positioning performance of particle filter is better than that of extended kalman filter.Next,a simulation platform was built based on ROS(Robot Operating System).A 3d modeling software was used to build an AGV model with four-wheel differential steering and Ackermann steering at the same time,which was imported into the simulation platform.The AGV path accuracy of different steering modes was compared through simulation,and the mapping effect of SLAM algorithm based on particle filter and SLAM algorithm based on graph optimization was compared.Through calculation,motor and lidar were selected,an AGV was designed and built,and the AGV's autonomous navigation function and SLAM algorithm were tested.Finally,the simulation platform was used to provide data sets to complete the optimization of SLAM parameters.Levenberg-marquardt method was used to improve the back-end of Karto SLAM.