Research On AGV Navigation And Positioning Technology Based On Visual Neuron

Aiming at the problem of low navigation and positioning accuracy in the navigation and positioning process of vision Automated Guided Vehicle(AGV),this paper proposes a vision guided AGV based on neuron processing system for navigation and positioning.The visual neuron processing system includes H-H neurons for signal channel allocation,FHN neurons for subsequent processing of ultrasonic signals,and RBF neurons for setting optimal weights in data fusion.Neuron system is used to process multi-sensor integrated navigation and positioning data,so as to assist visually guided AGV to improve navigation and positioning accuracy.Therefore,the application of neuron processing system in the vision guided AGV navigation and positioning system in this paper can well improve the navigation and positioning accuracy of AGV,and has important scientific significance and practical value for the practical research of AGV navigation and positioning technology.Specific work is as follows:Firstly,the navigation accuracy of AGV is improved.Therefore,this paper establishes an image digitization system containing H-H neurons and sensor auxiliary data processing system,and corrects AGV driving direction by tracking controller with PID correction path control algorithm of visual neuron.In order to ensure that the AGV can correct its motion path more quickly and optimally,the mathematical model of brushless DC motor and the kinematics model of AGV are used to construct the optimal control function relationship.In addition,the deviation correction voltage is introduced into the dual-channel PID system to achieve real-time adjustment of the differential effect of the driving wheel,so that the AGV can realize a more rapid and optimized path for deviation correction.Secondly,in order to improve the positioning accuracy,the Horizontal Dilution Precision(HDOP)principle is adopted to determine the installation position of the ultrasonic sensor assisted by visual guidance from the installation position to the measurement data processing,and the geometric constraint theorem is used to screen out the high-precision combination positioning data.In order to improve the accuracy of pseudo-observations and reduce the computational amount of geometric constraint processing,considering the significant weak signal perception ability of FHN neuron system and the random-resonance phenomenon caused by noise can enhance the weak signal detection ability of the neuron system,an ultrasonic receiving sensor based on the random-resonance characteristics of FHN visual neuron system is proposed,so as to enhance the signal and denoise in real time.Improve positioning accuracy.Then,multi-sensor navigation and positioning data are fused to improve navigation and positioning accuracy.In the process of multi-sensor data fusion,RBF visual neuron system was used to set the optimal weight factor of neurons to solve the nonlinear relationship of observed data.According to the nonlinear relation of prediction equation treated by untraced transformation in adaptive untraced Kalman principle,the accuracy of sigma points can be improved by using the constraint condition of untraced point,so as to improve the accuracy of subsequent prediction,which can effectively ensure the advisability of prediction points in untraced transformation,and finally improve the precision of navigation and positioning.Finally,a vision guided AGV test platform was built,and the vision guided AGV dualchannel PID path control algorithm was tested,and the method of determining the installation ultrasonic position according to the plane geometric accuracy factor and the navigation and positioning data fusion algorithm were tested.