.png){kind=icon}
.png){kind=icon}
Return
Self-Organizing Edge Computing Distribution Framework for Visual SLAM
.png){kind=icon}
OA
AI
DOI：10.1109/LRA.2026.3679267.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Localization in a known environment is an essential capability for mobile robots. Simultaneous Localization and Mapping (SLAM) addresses this by combining real-time tracking with computation-intensive map optimization, which can present a challenge for resource-limited robots. Edge-assisted SLAM approaches that offload heavy computation while maintaining real-time tracking onboard offer a potential solution. In this article, we propose a novel self-organizing VSLAM framework that provides a general structure for distributing existing keyframe-based VSLAM systems across a network of devices. The framework introduces a state management model for handling shared SLAM state among devices and a distribution policy for orchestrating the distribution in a self-organizing manner. To demonstrate the framework, we implemented it for monocular ORB SLAM3 using a three-layer architecture. The distributed SLAM was evaluated in both fully distributed and standalone configurations and compared against the original ORB SLAM3. The experimental results show that the proposed framework achieves comparable accuracy and resource utilization to ORB SLAM3. Moreover, the system can revert to standalone SLAM when network connectivity is lost, demonstrating effective self-organization.
Keywords:
SLAM
distributed robot systems
software architecture for robotic and automation
Journal
I
IF:
5.3
Papers:
1.7K
Citations:
3.9W