*4.3. Knowledge Base*

The ability to share knowledge is a crucial prerequisite for teamwork. As discussed in Section 3.6, there are various approaches for building a shared knowledge base in multi-robot systems. Some standardization would also be helpful here. Important research questions to ask in this realm concern the integration of heterogeneous knowledge representations, the implementation of the knowledge base in a distributed system with largely diverging agent capabilities, the satisfaction of stringent application performance requirements, the extensibility of the knowledge base structure at run-time, i.e., adding new concepts and relations, as well as removing invalid facts, and the inclusion of "common-sense knowledge" that humans would have implicitly, but which needs to be provided explicitly to a robot team.

For all of these aspects, the state of the art provides individual solutions. However, how to forge these solutions into a shared knowledge base that satisfies the specific requirements of an application domain is an open problem. Moreover, we need to explore whether it is feasible to reduce the large number of approaches to a few consolidated ones.
