A common data fusion framework for space robotics: architecture and data fusion methods

Data fusion algorithms provide a system with the capacity to combine the data from different sensors and metadata (e.g. timestamps, geometric models) into into symbolic representations like maps or position estimations. Software that implements these algorithms needs to provide a solution to the challenge of awareness using multiple sensors faced by robots and autonomous systems as well as the means to access and store this data for future uses. The software community in space robotics lacks a common framework to make developing, reusing and comparing data fusion solutions easier. InFuse provides that framework, and not only a set of data fusion solutions based on state-of-the-art algorithms, but also performance metrics to qualify algorithms. InFuse is developed by six industrial and academic partners working in the space sector, under the supervision of several European space agencies (ASI, CDTI, CNES, DLR, ESA, UK Space). This paper describes the architecture and the methods of the robot perception and localization framework InFuse Operational Grant 3 (OG3), in order to provide the Space Robotics community with a Common Data Fusion Framework (CDFF).