Controller Design And Path-Following Algorithm Research For AGV Based On Vision Navigation

AGV is a kind of industrial equipment which is used to carry material in different kinds of production lines. The stability and rapidity of running process of AGV have important influence on the efficiency of the whole line.And ability to adapt to different paths has effect on the flexibility of the manufacturing system. This article researches on AGV control system in order to solve the problems keeping AGV following the scheduled path steadily. It designs a controller based on vision navigation which can follow preset path smoothly and rectify the deviation timely. Besides, it can detect special landmarks, complete the default actions and follow the external instruction as well. The major research works are as follows:This paper makes an overall system plan according to project demand. The plan takes vision information from industrial camera as feedback and develop algorithm framework based on this method. It also proposes a method to set up particular actions for each special function points.Then this paper states how to get values of route parameters and contrast the effect of linear fitting with the effect pf circular arc fitting. After completing camera calibration, it can get the real error parameters and take them as the input of the path-following algorithm. Besides, this paper works out a method to calculate the similarity degree of the object and template in order to detect the landmarks.This paper builds a motion model with limited drive capability, and takes advantage of an existing algorithm, which aims at the optima coordination of deviation, to meet the path-following requirements. It finishes path-following process on a model with a side-installed camera in a situation with large-deviation at initial status.This paper modifies the algorithm to apply to path-following of large curvature radius. To work out the control output according to this new algorithm can help solve the problem that to follow a large-curvature circular path with smaller error comparing to the old algorithm at the same running speed. To solve the problem that the length of time to prediction and control both need to debug, this paper increases the number of predict steps, and reduce scope of possible control output of each step, meanwhile, the calculation amount of each step remain the same.This paper introduces some modules of the controller hardware and software which are used to deal with the external instructions and internal logic of actions. A experimental platform is built, and the experiment results verify the practicality of the new algorithm and obtain a good effect which confirm that the algorithm proposed in this article can help get smaller path-following error than the existing algorithm.