Design And Implementation Of AGV Visual Localization System

With the fast development of intelligent logistics and automatic warehouse,Automated Guided Vehicle(AGV)has been widely used as a key component of those systems.However,the state-of-art localization technology of AGV can't meet the requirements for accuracy and expansibility.As a solution of robotic localization,Simultaneous Localization and Mapping(SLAM)has drawn significant attention.In recent years,visual SLAM has become a research focus due to its unique advantages..Aiming to solve the localization problem of AGV with visual SLAM,this dissertation designs a visual localization system which can work stably in complex factory environment.The research presented in this dissertation is based on the project "Visual Guided Automated Forklift system".It is implemented through theoretical analyses and practical experiments with respect to the visual localization system of AGV.The main research in this dissertation are described as follows:1.In order to improve stability and practicability of SLAM system in the industrial,this dissertation designs a visual SLAM algorithm based on artificial beacon.The artificial beacon is designed based on hamming code and its corresponding recognition algorithm is proposed.They can improve the reliability of data by self-checking.On basis of the beacon,a new localization algorithm and graph optimization model are proposed to improve the localization accuracy.2.For improving the artificial beacon based SLAM algorithm,a multi-view SLAM algorithm is put forward.It includes a stereo match algorithm,a pose tracking algorithm in the front-end and a global optimization algorithm in the back-end.In stereo match algorithm,the matched sub-pixel point pairs are obtained by SAD matching and parabola fitting.In pose tracking algorithm,we implement a stereo visual odometry through RansacPnP and local optimization.In the back-end,we propose a new optimization algorithm based on global planar constraint and artificial beacon closed-loop detection.3?The design and implementation of AGV system,including the hardware and software architecture of both terminal control system and vehicle-mounted control system.Terminal control system consists of map management system,task scheduling system,path planning system and trajectory interpolation system.Vehicle-mounted control system includes localization system,driving system,power management system and obstacle avoidance system.Data transmission between terminal control system and vehicle-mounted control system is realized via wireless communication.4?NVIDIA embedded controller--Jetson TX1 is used in our system.We have conducted the simulation analyses and practical experiments in the factory.The results verified the accuracy and stability of this proposed system.