Abstract
Visual simultaneous localization and mapping (vSLAM) is a prevailing technology for many emerging robotic applications. However, real-time SLAM on mobile robotic systems with limited computational resources is hard to achieve because the complexity of vSLAM algorithms increases over time. This restriction can be lifted by offloading computation to edge servers that have abundant computational resources, forming the emerging paradigm of edge-assisted SLAM. Nevertheless, sending high-dimensional vision data over volatile wireless channels inevitably leads to excessive delay, which decelerates map updating and thus degrades the localization accuracy. In this paper, we design a new system architecture that enables low-latency communication for edge-assisted SLAM with improved localization performance. Extensive experiments prove that our system is cost-effective and outperforms better.