Research And Implementation Of Robot Location And Navigation Algorithm Based On Convolutional Neural Network And Depth Camera

Usually,the classical processes of feature extraction,feature matching and tracking are followed in the localization and navigation of robots,and other optimization strategies are used to optimize them.However,in the face of complex scenarios,such methods often have the shortcomings of insufficient stability,and many steps and complex processes.Deep learning technology shows strong feature abstraction and expression ability when dealing with image understanding tasks,which provides a new research idea for robot positioning and navigation.Therefore,this paper studies the problem of robot positioning and navigation.(1)Combining the nature of tasks and the characteristics of in-depth learning,the problem of robot positioning and navigation is modeled as a supervised regression learning problem,and the overall algorithm framework is analyzed and designed.(2)A robot localization model based on convolution neural network is proposed.This method uses convolutional neural network with image frame input to complete feature extraction,uses shared full connection layer and loss function to directly predict inter-frame pose vectors,and then constructs a pose regression model.This method not only simplifies the process of feature extraction,but also integrates different levels of features and improves the prediction accuracy of the model.(3)The robot navigation method based on A*algorithm is completed.The algorithm takes the location information output from the robot positioning network as input,completes the static path planning,has fast speed and high real-time performance,and meets the needs of robot positioning and navigation.In summary,this paper uses depth neural network to design and implement an improved method of robot positioning and navigation based on depth camera step by step,and designs experimental tests.The experimental results show that the method has good effect on various performance indicators and has practical application value.