Research On Automatic Guided Vehicle System Based On Visual SLAM

With the rapid development of artificial intelligence,the automated guided vehicle system has a very high market prospect and research value.At present,the automated guided vehicle system is still dominated by laser navigation,so it needs high construction costs.With the continuous improvement of the level of camera sensors and the development of machine vision technology,low-cost camera sensors can also complete the autonomous mapping and navigation tasks of robots,but there is still a big gap compared with the accuracy of laser radar,so developing a low-cost and reliable automated guided vehicle system has strong industrial application value.In order to solve the above problems,this paper studies and designs a low-cost Mecanum wheel automated guided vehicle system based on visual SLAM,and deploys the algorithm on the embedded development board.It collects environmental information through the depth camera to build a 3D map,and finally realizes autonomous navigation in the indoor environment.The specific research is divided into the following aspects:First of all,on the selection of visual SLAM algorithm,the appropriate algorithm is selected based on the scene requirements and embedded carrying conditions.Based on this algorithm,the IMU,wheel encoder and visual odometer are fused through data fusion to improve the positioning accuracy of the automated guided vehicle,so as to optimize the mapping effect and ensure reliable mapping for subsequent navigation work.Secondly,in the embedded implementation phase,compared to the PC platform,the computing power level of the embedded platform is relatively limited.This paper combines the characteristics of NVDIA Jetson Nano,which has both CPU and GPU computing capabilities,to accelerate the feature extraction and matching phase of the visual SLAM algorithm through CUDA,improving the real-time performance of the algorithm running on the embedded development board.Compared to the original RTAB-Map algorithm,the average tracking time has been improved by 10.42%.Moreover,in the selection of path planning algorithm,select the global path planning algorithm and local path planning algorithm suitable for the automated guided vehicle model,and optimize and adjust the related calculation amount,path smoothing adjustment,and algorithm related parameters based on the algorithm to ensure the smooth movement of the automated guided vehicle system.Finally,in terms of system implementation,NVDIA Jetson Nano is selected as the ROS main control,STM32F4 is used as the lower computer to drive the guided vehicle,and Wi-Fi wireless communication technology is used to achieve remote PC control.The results show that the indoor environment map constructed by this system can provide reliable mapping for the subsequent navigation environment and has good real-time performance.In the navigation link,the automated guided vehicle can also plan a feasible route according to the path planning algorithm and move forward smoothly according to this route.It is a low-cost automated guided vehicle system and has good application value.