Building Complex Systems with Agent-Space Architecture

Abstract

Building complex systems requires a specific kind of modularity as well as incremental development. Here we introduce an architecture where basic modules are reactive agents and the data exchange among them is based on the so-called stigmergic communication through space. In this way we have connected ideas coming from the multi-agent systems and the coordination languages on the one side and ideas of the behavior-based systems on the other side. We demonstrate that this architecture manifests several interesting features which are useful for engineering of real-time systems and modeling of biological creatures or their parts. We advocate for so-called purely reactive agents which are stateless entities usually taken as too weak building blocks of systems. However, their features enable us to use a special method of incremental development (so-called subsumption method).