Architecture and Application Abstractions for Multi-Agent Collaboration Projects

In this paper, we propose that allapplication for multi-agent collaborative control, and in particular fixed-wing umanned arial vehicles, can be be described through three basic behaviors. They are Travelling, Wactching, and Tracking. These three behaviors are constructed from two primitive actions, going somewhere and hlding at a particular location. The semantics of the behaviors are clearly difined and justify why these three are chosen. The ability of these behaviors to easily program different missions through hyvrid state flow is illustrated through examples of previously-demonstrated application. Furthermore, this paper outlines three levels of abstraction for a software architecture for collaborative control. In particular, we focus on an procedural-level abstraction that illustrates where the above three behaviors can be used to define applications. This level of abstraction displays software modularity such that different members of a research group can work independently while maintaining a structure that allows successful experimental implementation on the multi-agent platform. Moreover, this modularity allows project members to perform various disparate applications with the platform without considerable software modification. It also promotes software reuse after members leave the group. While focusing primarily on fixed-wing aerial vehicles, the approach presented here is equally applicable to ground and rotorcraft vehicles.