Research On AGV Location Method Based On Multi-sensor Data Fusion

AGV(Automated Guided Vehicle)is an integrated equipment that integrates mechanical,control,optical,electronic,computer,and Internet of Things technologies.Because of its intelligent operation mode,flexible path management,efficient space utilization and many other advantages,it is widely used in the automotive industry,military manufacturing,warehousing and logistics,home appliance production and other industries.Location technology is the prerequisite for the automation and intelligent operation of AGV,and it is also one of the most important standards to measure the development level of AGV technology.The main contents of this paper are as follows:(1)This paper introduces the key technologies in the field of AGV location,and completes the construction of the location system research platform,including the definition of the coordinate system,the construction of distributed network architecture,the acquisition and processing of sensor data and so on.(2)This paper studies and analyzes a variety of existing AGV location methods,and clarifies the basic idea of how to use multi-sensor data to optimize the AGV location effect,and proposes an AGV location method based on laser radar,camera and IMU(Inertial Measurement Unit)attitude sensor data fusion.In this method,aiming at the problem of insufficient location accuracy of lidar,the camera is introduced to achieve higher precision environment perception,and the camera-based marker recognition is used for location,the computing resource consumption of location system is controlled well.Aiming at the limited range of high-precision marker recognition,the location result obtained by the lidar scan matching is used to assist in determining the accurate pose of the marker currently recognized by the camera,and then the high-precision AGV location results are obtained.The size of the marker identified by the camera is used to judge the reliability of the location result of the camera mark recognition.The analysis and test results show that the multi-sensor data fusion localization method proposed in this paper has better AGV location effect under certain conditions.(3)This paper attempts to apply the neural network to the AGV location system,and designs a multi-sensor fusion location model based on BP neural network,the model takes the lidar and IMU sensor data as input and the AGV pose as the output,the multi-sensor fusion location is realized by the self-learning ability of BP neural network.Aiming at the continuous trajectory of AGV,a method of applying the cyclic neural network to the AGV location system is proposed.The experimental results show that the BP neural network is effective for AGV location,and the pose prediction results of the cyclic neural network during the location process are also accurate.