A Planning Pipeline for Large Multi-Agent Missions

Abstract

In complex multi-agent applications, human operators are often tasked with planning and managing large heterogeneous teams of humans and autonomous vehicles. Although the use of these autonomous vehicles broadens the scope of meaningful applications, many of their systems remain unintuitive and difficult to master for human operators whose expertise lies in the application domain and not at the platform level. Current research focuses on the development of individual capabilities necessary to plan multi-agent missions of this scope, placing little emphasis on the integration of these components in to a full pipeline. The work presented in this paper presents a complete and user-agnostic planning pipeline for large multi-agent missions known as the HOLII GRAILLE. The system takes a holistic approach to mission planning by integrating capabilities in human machine interaction, flight path generation, and validation and verification. Components - modules - of the pipeline are explored on an individual level, as well as their integration into a whole system. Lastly, implications for future mission planning are discussed.