Abstract: Vision-based multi-robot global localization in large-scale environments is a challenging task ... primarily due to the limited descriptive power of feature information (semantic or ...