| The improvement of Industrial automation and intelligence level puts forward higher requirements in the logistics system. Automatic Guided Vehicle(AGV) as an important part of modern logistics system, has been widely used. Among ways of AGV navigation, visual navigation has a greater advantage in the path of change, intelligent obstacle avoidance, and functional expansion. And it has broad prospects for development as well. This paper researches for AGV vision navigation technology, and the main contents are as follows:Firstly, on the basis of visual development status and key technology of AGV visual navigation, the focus of this paper has been determined to be localization technology and path planning. Lots of aspects have been compared and analyzed to fix route navigation and the free path visual navigation. Herein, visual navigation scheme has been determined and hardware has been selected and commissioned.Secondly, visual localization algorithm has been studied. Based on visual processes, performances of various featured matching algorithms have been compared and analyzed. Consequently, ORB has been determined to a matched algorithm. In addition, Visual Dictionary matching algorithm has been used to improve image matching speed. Bayesian filtering algorithm has been used to establish a probability model to improve the image matching robustness. According to consistency of random sampling and least squares method, the position and orientation between matched images have been calculated. By experiment, the visual localization algorithm has been evaluated.Thirdly, based on AGV free path navigation, the research has been carried out in two ways: global path planning and local path planning. To get a path with shortest running time, an improved JPS(Jump Point Search) global path planning algorithm has been proposed based on omnidirectional mobile node culling method, with cost evaluation method combined with the use of angle. At the same time, to solve the problem of local minima, the policy along the wall has been introduced to propose an improved dynamic window local path planning algorithm. The simulation results has demonstrated its feasibility of the two algorithms.Finally, AGV vision navigation experimental platform has been built, which has been applied to complex industrial sites. The results has showed that visual localization algorithm and path planning algorithm proposed in this paper have higher accuracy and robustness. Thus, the visual navigation system based on visual localization algorithm and path planning algorithm proposed in this paper can be used in the implement of AGV autonomous navigation. |
PDF Full Text Request