Research And Implementation Of Web SLAM Positioning Technology Based On Device-Cloud Collaboration

SLAM technology has developed rapidly and has been widely used.Web SLAM is regarded as a new growth point for SLAM applications due to its advantages of light weight,easy sharing,and low barriers to use.However,limited by the Web environment,there is still a lot of room for improvement in the actual application effect of Web SLAM.From the perspective of device-cloud collaboration,this thesis focuses on solving the technical pain points in the positioning process of Web SLAM.The main work content of this thesis is as follows:(1)Aiming at the contradiction between the local computing power and the environment of Web SLAM,which is difficult to meet the needs of complex positioning-algorithms,a device-cloud collaboration architecture suitable for Web SLAM is proposed.Based on the existing SLAM terminal-cloud collaboration architecture,a pose prediction and pose fusion scheme is proposed for the actual cloud delay problem.(2)Aiming at the long-distance problem existing in the existing deep learning visual-inertial odometry,a Transformer-based visual-inertial state estimation-VIformer is proposed.Using Transformer’s long-distance dependence feature to improve the performance of deep learning odometry in long-distance problems.Compared with other deep learning mileage calculation methods,VIformer’s pose estimation percentage error decreases as the distance increases.(3)For the integration of Web SLAM systems.Using the device-cloud collaboration architecture,the existing lightweight Web SLAM platform is extended to a device-cloud collaboration Web SLAM system.It not only retains the real-time advantages of the lightweight Web SLAM platform,but also comprehensively improves its positioning accuracy.Based on the above work content,this thesis implements a Web SLAM positioning system based on device-cloud collaboration,and conducts related tests.The results show that the system realizes the positioning function,and compared with the lightweight Web SLAM platform,the positioning accuracy is improved.A new implementation scheme is provided for the Web SLAM localization problem.