Using a Genetic Algorithm to Evolve a D* Search Heuristic.
Andrew Giese, Jennifer Seitzer · 2011
Evolutionary computation (EC) is the sub-discipline of artificial intelligence that iteratively derives solutions using techniques from genetics. In this work, we present a genetic algorithm that evolves a heuristic static evaluation function (SEF) function to be used in a real-time search navigation scheme of an autonomous agent. This coupling of algorithmic techniques (GAs with real time search by autonomous agents) makes for interesting formalistic and implementation challenges. Genetic evolution implies the need for a fitness function to guide a convergence in the solution being created. Thus, as part of this work, we present a fitness function that dictates the efficacy of a generated static evaluation function. In this work, we present algorithmic and formalistic designs, implementation details, and performance results of this multi-layered software endeavor.