Research on Human-Robot Collaborative Mapping System Based on Visual SLAM

Abstract

Nowadays, with the widespread application of SLAM technology, mapping has become one of the important application directions. However, traditional single machine SLAM (Simultaneous Localization and Mapping) systems typically only use a single ground or aerial device for mapping. Ground robots have limited perspectives and data collection ranges, resulting in low efficiency. In contrast, aerial robots have the advantage of a wide perspective, enabling them to quickly perceive the environment over large areas. To address this issue, this study proposes a human-machine collaborative mapping system. In this system, an innovative approach is proposed to use ground end human handheld smartphones as mapping devices to collaborate with ground mobile robots. At the same time, aerial drones can compensate for the shortcomings of ground devices by collaborating with handheld devices and ground mobile robots for mapping. Each device extracts image information from the spatial environment through monocular vision sensors, estimates pose through feature point matching, and creates local maps separately. In addition, the system effectively improves the problem of map fusion failure between multiple devices by constructing a keyframe database for multi map matching. The authenticity of the system has been verified in actual campus environments. The experimental results show that the method can not only fuse local maps from different devices, but also improve the pose accuracy after fusion, effectively improving map coverage and trajectory accuracy.