Joint simultaneous localization and mapping (SLAM) constitutes the basis for cooperative action in multi-robot teams. We designed a stereo vision-based 6D SLAM system combining local and global methods to benefit from their particular advantages: (1) Decoupled local reference filters on each robot for real-time, longterm stable state estimation required for stabilization, control and fast obstacle avoidance; (2) Online graph optimization with a novel graph topology and intra-as well as inter-robot loop closures through an improved submap matching method to provide global multi-robot pose and map estimates; (3) Distribution of the processing of high-frequency and high-bandwidth measurements enabling the exchange of aggregated and thus compacted map data. As a result, we gain robustness with respect to communication losses between robots. We evaluated our improved map matcher on simulated and real-world datasets and present our full system in five realworld multi-robot experiments in areas of up 3,000 m 2 (bounding box), including visual robot detections and submap matches as loop-closure constraints. Further, we demonstrate its application to autonomous multi-robot exploration in a challenging rough-terrain environment at a Moon-analogue site located on a volcano. K E Y W O R D S graph SLAM, map matching, mobile robots, multi-robot, navigation filter 1 | INTRODUCTION The exploration of moons and foreign planets is an important current and future application for mobile robots as their surfaces are difficult to reach and hard to access for humans. The application of huge and complex robot systems such as Curiosity, landed on Mars in 2012, creates many single points of failure for a mission. As a consequence, these rovers have to move very slowly and carefully to avoid getting stuck, as the Mars rover Spirit did in 2009 (Wolchover, 2011). The future deployment of teams of multiple robots can avoid these single points of failure by gaining robustness through redundancy and, in addition, can improve efficiency through parallelization. The robots have to travel through previously unknown unstructured rough terrain, operating in areas where external methods for localization like global navigation satellite systems (GNSS) are not available or expensive to set up. Communication links to the robots are limited and heavily delayed, featuring for example 8-40 min round trip time between Earth andMars. Furthermore, communication between the robots cannot be guaranteed at all times, in particular at scientifically interesting places such as craters, canyons, or caves. As teleoperation therefore becomes inefficient or infeasible, robot autonomy is a key aspect for future planetary exploration missions. Any coordinated (semi-)autonomous operation in such challenging environments requires up-to-date localization estimates for all robots in a team as well as a joint map to operate on.