I. Introduction
SLAM (Simultaneous Localization And Mapping) has been widely studied in the fields of robotics and computer vision. The robustness and accuracy of a SLAM method in unknown dynamic environments can hardly be achieved using only one sensor. The paradigm between computational power and availability of new sensors have lead to obtain large-scale dense maps by developing Multi-SLAM methods that cooperate to improve localization and to speed up registration of a detailed 3D representation of the environment. Applications such as autonomous multi-robot (MR-SLAM) navigation [1]–[3], parallel indoor/outdoor 3D registration [4], and multi-session mapping [5], [6] are very active fields in the literature.