Using collaborative sharing on cloud for fast relocalization in keyframe-based SLAM

Abstract

Relocalization when tracking fails during simultaneous localization and mapping (SLAM) is still a task full of challenges, especially for these using keyframe technology to reduce backend pose graph size. These challenges come from two aspects, which include lack of enough image data when tracking fails and the high computational complexity which can't afford by local robot. In this situation, even the state-of-the-art keyframe-based SLAM may not be fast enough to recovery from the tracking failure state. However, the emerging cloud robotics has enlightened a new direction to address relocalization problem in both two factors and in this paper we present an approach based on cloud-based sharing, which aims at providing a way for fast relocalization on the existing keyframe-based SLAM framework. Our method can effectively utilize the sharing environment map data contributed by large scale of robots for the local relocalization and also proposes various mechanisms to eliminate the degeneration of this distributed model and the unstable network. We have realized the prototype, and made it cooperation with the leading framework, ORB-SLAM. The evaluation results also show that our method does have the ability for fast relocate itself compared with the original setup and retains the high efficiency of the original SLAM framework in normal state.