Using Genetic Algorithms to Represent Higher-Level Planning in Simulation Models of Conflict

The focus of war f are h as shifte d f rom t he In dust r i al Age to t he In for m ation Age, as encapsulate d by t he ter m Ne twor k E nabledCapabilit y. T his emphasises infor mation shar ing , command decision-making , a nd the resultant plans m ade by commanders onthe b asis of that infor mation. P lanning by a hig her l e vel militar y commander is, i n m ost cases, regarded as such a difficult processto emulate, that it is per for med by a real commander dur ing wargaming or dur ing an ex per i mental session based on a Sy ntheticEnv i ronment. Su ch an approach g ive s a r i ch representation of a s mall nu mber of data points. Howe ver, a m ore complete a nalysisshould allow s earch acro ss a w ider set of alter natives. This re quires a closed-for m version of such a s imulation. In this paper, wediscuss an approach to this problem, based on emu lating the hig her command process u sing a combination of game theor y andgenetic algor ithms. This process was initially i mplemented in an ex plor ator y research initiative, descr ibed here, and now for ms thebasis of t he de ve lopment of a “Mission Planner,” p otentially applicable to all of o ur hig her l e vel closed-for m s imulation m odels.