No abstract

We examine the problem of self-adaptation in game-playing agents as the game requirements evolve incrementally. The goal of our current work is to develop an interactive environment in which the game designer generates requirements for a new version of a game, and the legacy software agents from previous versions of the game adapt themselves to the new game requirements. We are developing and testing our metareasoning technique for adapting a game-playing agents in Freeciv, a mature program in the domain of turn-based, multi-player strategy games. In this paper, we first present an analysis of adaptations to FreeCiv, next describe our general approach, and then describe a specific adaptation scenario.