A Robust Client-Server Architecture for Map Information Processing and Transmission for Distributed Visual SLAM

This paper presents the development of a framework for distributed processing and transmission of key frames and map points, which are components of a sparse map built with feature based visual SLAM frameworks. The developed framework uses a client-server architecture that makes it possible to separate the map and key frame components of the visual SLAM for communication in real-time. We devise a mechanism to minimize the communication bottleneck that is encountered when transmitting map points. The implementation of transmitting map is generally a bandwidth intensive process. The designed architecture provides excellent overall map quality even in scenarios where map objects are lost in the communication process. We perform simulation based verifications and real world examples to illustrate the validity and robustness of our architecture. Our architecture provides the first such system to process and transmit full map information from client to server in real-time without sacrificing robustness.