Vision-based Automatic Tracking AGV Control System Design

Using AGV can greatly improve the efficiency of enterprise automation productionlines and logistics systems, and improve the level of automation of enterprises, so rapidlydevelopping AGV has become an inevitable trend. In recent years,with the development ofcomputer technology and image processing technology, Vision-based navigation method isincreasingly becoming a hot research issue due to its simple path setting,low costs,easymaintenance,good flexibility and so on.In this paper,Using a automatic tracing AGVvehichle based on monoculer in lobaratory as research subject, research works mainlyinclude identification and measurement of the guide track and control strategies.And theseworks are summarized as follows:Firstly，preprocessing for images captured from the visual sensor,and Extracting thecenterline of trajectory.For traditional detection algorithm which transform all types oftrajectory models approximately into linear model by hough transform is difficult to meetthe accuracy requirements of the curve trajectory,here a new method is presented oftrajectory model classification based on average slope difference and conerity distinguishindex according to local guide features by visual inspection which divides traiectory intoseveral types,Using MMSE and Levenberg-Maquardt method to fit linear and arc,etc.model respectively, Ultimately depending on the guide parameters given by differentmodel the navigation accuracy of vision-guided AGV is improved.Secondly,in the control strategy optimization,for the eye-in-hand robot visual servoingsystem,a visual controller is designed using the image error and visual servoing modelcombined with CLF(control Lyapunov function) method. Using the output of visualcontroller as translational and rotational velocity of the robot, and then a dynamicscontroller is designed based computed torque method.Realizing the point stabilization taskbased on image,while solving the problem of Velocity mutation of AGV under control bya single visual controller;A path tracking controller is designed based on AGV’s visualservoing model combined with fuzzy control theory and the theory of variable structurecontrol(VSC).Using the error detected in the image plane as input of the fuzzy controller to realize tracking tasks for several typical continuous and discrete trajectory. Comparedwith traditional variable structure control method,this method has the following featuresinclude tracking the trajectory without overshoot,and converging quickly in big deviationconditions.Finally, the trajectory recognition and measurement algorithms and the controllerdesigned in this paper are simulated by MATLAB/Simulink, through objective analysis ofthe experimental results verify its correctness. The above algorithms are realized throughvc6.0sofeware programming,and conducted experiments on the AGVsystem.Experimental results demonstrate the effectiveness and practicality of the designscheme.