Research On Composite Navigation Technology Of Intelligent AGV Based On Multi-Vision And Inertial Navigation

Compared with other external measuring sensors,a large amount of perceptual information such as color,shape and texture is provided by vision guidance,which is beneficial to improve the accuracy and intelligence of the AGV navigation system.However,until Amazon introduced the warehousing logistics robot KIVA,the vision recognition and loctation technology was combined with the two-dimensional code landmark,which prompted the vision guidance method to be widely used in industry.In this context,the industry hopes to extend the two-dimensional code vision navigation technology to the manufacturing production line,but the complex industrial environment brings many difficulties to the application of two-dimensional code landmarks.For production line scenarios with material warehouses,firstly,this paper analyzes the different requirements of the machine tool production line and material warehouse for the accuracy,speed and reliability of AGV positioning navigation.A composite navigation scheme combining multi-camera switching,landmark recognition and ontology sensing,ribbon tracking and two-dimensional code positioning is designed.The AGV's composite navigation technology with fusion ribbon guidance,two-dimensional code positioning and INS estimation is studied.Secondly,in the machine tool production line with large ground vision interference,high positioning accuracy requirements and limited loading and unloading space,a multi-camera vision composite navigation scheme using topological map,ribbon tracking and two-dimensional code loctation is adopted.The method of extracting the features of the stained ribbon,the path fitting method of the ribbon target point,and the method of the standard two-dimensional code identification and measurement are studied.Based on the research of the above methods,a fast two-dimensional code region localization and recognition measurement method based on ribbon features is proposed.The ribbon and two-dimensional code recognition measurement experiments were performed on a high-speed moving conveyor belt,which measured the position and posture of the ribbon and the two-dimensional code in the camera field of view and verifies the effectiveness of the proposed fast two-dimensional code identification measurement method.Thirdly,in the material warehouse with large work scene,high shelf density and unmanned level,the composite navigation scheme with grid map,two-dimensional code positioning and INS estimation is adopted.At the same time,a real-time estimation method of AGV pose based on unscented Kalman filter is proposed.On the virtual path between two-dimensional code landmarks,continuous pose estimates are obtained by fusing estimates of the kinematics model and observations of the inertial navigation system.At the landmark location,by measuring the global position and attitude of the two-dimensional code,the cumulative error of the continuous pose estimation value is eliminated,and the accuracy and flexibility of the composite navigation method are improved.Finally,for production line scenarios with material warehouses,an AGV composite navigation system that combines ribbon guidance,two-dimensional code positioning,and INS estimation is developed.At the machine tool production line,the running test of multi-camera vision composite navigation of ribbon tracking and two-dimensional code positioning is completed.The developed vision navigation AGV has been applied to the actual factory.In the laboratory environment,the composite navigation experiment of two-dimensional code positioning and INS estimation is completed,and the effectiveness of the proposed composite navigation technology is verified.