In the ocean and space exploration application domains, robots require a high level of on-board autonomy based on a local interpretation of sensor data. The goal of this work page is to develop methods for on-board processing of sensor data and online navigation for collaborative localization and mapping with heterogeneous teams of robots. The resulting maps encompass thematic data relevant for the applications in both domains and data required for on-board navigation purposes, in particular to enable robot autonomy in environments where no global external positioning system such as GPS is available. The methods should be distributed in order to share the workload between robots and avoid single points of failure. As sensor data is always subject to noise and measurement errors, it is crucial to model and estimate the respective uncertainties and process the data within a probabilistic framework. This requires trade-offs between estimation accuracy, computational efficiency and the bandwidth needed to exchange data between robots and to a ground station.