Rapid and robust large-scale SLAM mapping

Determine algorithmic and system-level approaches for Simultaneous Localization and Mapping (SLAM) that enable rapid and robust mapping of large-scale environments, overcoming the limitations observed in single-agent SLAM systems.

Background

The paper motivates multi-robot collaborative SLAM by noting that single-agent SLAM, despite significant progress, struggles to efficiently and reliably map large-scale environments. This limitation is particularly evident in complex real-world scenarios such as urban driving and large-area exploration where scalability, robustness, and global consistency are crucial.

The survey positions 3D Gaussian Splatting (3DGS) as a promising explicit representation for high-fidelity, real-time mapping, and discusses how collaborative multi-agent systems could address the scalability and robustness challenges. However, the fundamental problem of achieving rapid and robust large-scale mapping remains open, motivating research into architectures, data fusion, communication strategies, and consistency mechanisms.