A Centralized Cooperative SLAM System for Improving Positioning and Perception Accuracy in GPS-Denied Environments

One of the key points in robotic collaboration is localization. In some scenarios where the Global Navigation Satellite System (GNSS) signal is unavailable, robots must rely on their own sensors for environment perception and self-localization. Simultaneous Localization And Mapping (SLAM) technology, as a widely used method for perceiving the external environment, has been extensively applied in the field of robotics. The existing work faced problems with scalability, robustness, and data association inaccuracies, and therefore this paper developed a collaborative vision SLAM algorithm for multi-robots. The proposed system, based on a centralized collaborative architecture, adopts a centralized collaborative architecture, including a central node and several sub-nodes. We also conducted simulations and experiments on homogeneous and heterogeneous platforms, demonstrating that the system can perform high-efficiency and high-precision positional perception in indoor and outdoor environments.

I. Introduction

With advances in robot technology, various ground and aerial unmanned platforms have been widely used in military, disaster or wild search and rescue work. The clustering of unmanned systems is a developing trend, expanding the coverage of unmanned platforms and improve mission efficiency to a certain extent. Efficient and accurate positioning information and environmental perception information are the cornerstones of autonomous collaboration of unmanned swarms, directly determining the overall effectiveness and practical application effects.