Evolvable Recovery Membranes in Self-monitoring Aerospace Vehicles

In this paper we proposed and verified a methodology underlying the design of localised algorithms for complex multi-agent systems, exemplified by self-monitoring aerospace vehicles. In particular, we considered the emergence of selforganising impact boundaries and recovery membranes, separating damaged and/or potentially recovering regions from less affected agents. In order to identify phase transitions in system's dynamics, we investigated graph-theoretic and information-theoretic metrics, and incorporated them within fitness functions for a genetic algorithm. The GA involved a generation gap strategy and targeted a response time as well as spatial connectivity, temporal persistence and size of emergent boundaries and membranes. A variety of stable spatiotemporal patterns were produced under selection pressure, highlighting the potential for the design at the edge of chaos.