Project title: Distributed data-driven control of networked multi-agent systems with guarantees (DATA-MAS)
Host Institution: Technical University of Munich (TUM)
Host Supervisor: Prof. Sandra Hirche
Co-host Institution: École Polytechnique Fédérale de Lausanne (EPFL)
Co-host Supervisor: Prof. Colin N. Jones
Summary project: In the near future, teams of underwater robots will collaboratively collect waste from seafloor (currently performed by human divers) and restore healthy oceans. Similar multi-agent concepts can be extended to unmanned aerial vehicles or smart home devices, which lead to applications in environmental monitoring or internet of things. Multi-agent systems can achieve complex tasks that are difficult or impossible for single agents to solve. However, current control methods are very conservative as they require either known accurate agent models or enormous amounts of high-quality data. This is highly impractical, as (i) agents are becoming increasingly complex, which makes accurate models extremely difficult to obtain; and (ii) enormous amounts of data lead to expensive computation, which significantly slows down agent operation. To fully exploit the potential of multi-agent systems, control methods with efficient data use have to be developed. The objective of this project is therefore to develop optimal online distributed data-driven control algorithms for networked multi-agent systems with theoretical guarantees that use agent data much more efficiently than the existing solutions.