An Evolutionary Agent Approach to Dots-and-Boxes

T. Bossomaier, A. Knittel, M. Harré, and A. Snyder (Australia)

Keywords

evolutionary economic multi-agent systems dots boxes

Abstract

We describe the evolution of intelligent agent teams for the combinatorial difficult game of Dots and Boxes. Agents operate within an artificial economy deriving reward only from won games, with no other kinds of feedback. By pro viding representations of component graph structures, upon which agents can evolve rules, the agent teams achieve significant performance against existing AI players based around established mathematical frameworks.

Important Links:



Go Back