Research On Simultaneous Localization And Mapping Of AGV Based On Multi-sensor Information Fusion

With the continuous development of China's industrial modernization,the warehousing and logistics industry has transformed from a simple labor-intensive industry to automation.The requirements for automated guided vehicles(AGV)in the production process are also increasing.Intelligent autonomous Navigation is the development of AGV.AGV uses the onboard sensor data to create an environment map,automatically searches for the best route based on the mission goal,and finally reaches the destination to continue to complete other tasks.In this process,simultaneous localization and mapping(SLAM)is a key step.This article takes TurtleBot3 Burger as a prototype,builds a hardware and software platform,and studies AGV SLAM based on multi-sensor information fusion.Aiming at the common cumulative error problem in SLAM problem,multi-sensor information fusion method is used to reduce the cumulative error.At the same time,the graph optimization theory is used as the SLAM algorithm framework to optimize the global error.Aiming at the problem of map quality evaluation,a method for quantitative analysis of maps is proposed to reduce the influence of subjective judgment on the evaluation of maps.Specifically,it mainly includes the following:1.Establish a robot coordinate system,transfer the information of each sensor of the robot to a unified coordinate system,establish an odometer motion model,a lidar observation model and an IMU motion model based on different sensors,and finally establish the raster map needed to create the map model.2.Research on SLAM algorithm based on multi-sensor information fusion.Introduce SLAM algorithm based on graph optimization theory,including graph construction process and optimization method,and propose an improved SLAM algorithm framework based on graph optimization theory.The key content of the graph optimization SLAM algorithm is studied: the method of establishing constraints and data associations in the front-end part,the back-end optimization method,and the loopback detection strategy.Finally,an odometer calibration method based on IMU data is proposed,which integrates magnetometer,accelerometer and gyroscope information to calibrate the odometer to reduce the cumulative error of the odometer and verify the effectiveness of the method through experiments.3.System experiment and map evaluation research.Aiming at the problem of map quality evaluation methods,a quantitative analysis method is proposed,which uses five indicators to comprehensively analyze the difference between the map and the ground truth map,and uses experimental data to analyze the maps generated by various algorithms.Reduce the impact of subjective judgment on map evaluation.