Visual Aided Navigation Of AGV Based On Deep Learning Method

With the automation trend of industrial,industrial manufacturing and logistics also encounters unprecedented opportunities and challenges.Traditional manual logistics can not meet the fast and accurate customer needs,and AGV(Automated Guided Vehicle)has become the best choice for logistics enterprises.In recent years,with the rapid development and wide application of artificial intelligence,especially deep learning theory,the application of deep learning method to AGV's visual navigation function has become a very popular research topic.Although currently vision navigation technology can not meet the requirements of industry navigation accuracy,this topic will use the excellent feature extraction ability of deep learning in the computer vision,combined with the flexible vision technology,design and implement AGV navigation operation in a specific scene as an aided module in the complete navigation process,improve the navigation task of AGV.The works of this subject mainly include: Firstly,sort out the characteristics and trends of AGV navigation technology,study the advantages and applications of deep learning methods,summarize the main application directions and typical application examples of current deep learning in visual navigation field,and analyze the assistant role of deep learning in AGV visual navigation.Then,according to the AGV visual navigation task embedded development,study the structure and characteristics of SSD_MobileNet,a lightweight deep learning model,propose a targeted embedded optimization scheme,and test on Raspberry Pi hardware environment.Then,based on the optimized deep learning model,propose the overall design scheme of AGV visual aided navigation,construct a camera ranging model to convert the two-dimensional coordinates into three-dimensional actual position and distance information.According to the design scheme,we acquire images,retrain the optimized object detection model,and complete the model revision,finally integrate the complete visual aided navigation function module into the ROS platform,completes the realization of AGV visual aided navigation module;finally use the constructed navigation module to test navigation function,and Analysis and summary of the results.