Research And Implementation Of Mobile Robot Autonomous Navigation Method Based On Deep Learning

In recent years,the working scenarios and functional requirements of mobile robots have shown a trend of diversification and complexity.The autonomous navigation function is the basis for the mobile robot to complete other complex functions,and it is one of the most indispensable functions.At present,the autonomous navigation function widely used in mobile robots is mainly realized by lidar,which cannot use information with rich features such as images and languages for navigation and obstacle avoidance.How to make mobile robots understand images and language,and use the visual and semantic information obtained by them for autonomous navigation has become an urgent problem to be solved.This paper studies the application of visual and semantic information in the autonomous navigation of mobile robots,and proposes a deep learning-based autonomous navigation method for mobile robots,which mainly includes two aspects of visual-language navigation(VLN)and visual obstacle avoidance.In the research of VLN,a VLN method based on multimodal features is proposed.Compared to previous work,the proposed VLN method utilizes the object detection information based on the idea of bag-of-words model and the hidden state information of navigation history based on the idea of recurrent neural network to enhance the cognitive ability of the model to the current environment and navigation process,and reduce the risk of model overfitting.In the research of visual obstacle avoidance,a visual obstacle avoidance method based on convolutional attention mechanism is proposed.Compared to previous work,the proposed visual obstacle avoidance method uses the convolution block attention module CBAM in the feature extraction process in order to increase the model’s ability to perceive obstacles,which improves the prediction performance of the model without increasing the complexity of the model.In addition to testing in the simulated environment,this paper designs a deep learningbased autonomous navigation system and deploys the proposed method on a real mobile robot platform for testing.The contributions of this paper mainly include the following aspects: 1)A VLN method based on multimodal features is proposed,which improves the navigation performance of the model.2)A visual obstacle avoidance method based on convolutional attention mechanism is proposed to improve the prediction performance of the model with little increase in model complexity.3)In addition to testing in the simulated environment,this paper designs a deep learning-based autonomous navigation system and deploys the proposed method on a real mobile robot platform for testing.