Adaptive combat AI in strategy games- An approach based on Dynamic Scripting

When we talk about computer learning in competitive games, we think of an opponent that can learn from its mistakes and use creative new tactics to counter our own. Dynamic Scripting is a method for letting the game AI do just that – and do it in a believable way. The setting for this work is the strategy game World in Conflict. The dynamic scripting agent will be learning how to utilize its units and their different abilities to win tactical battles against different opponents. This thesis aims to show how a list of actions, actions that humans also use to make up their game play, can be combined by the dynamic scripting agent into fully working tactics, capable of adapting to beat its opponents. It will also investigate how the knowledge of such an agent can be used to create a static, well-performing AI. The results clearly show that both the learning agent and the static agent derived from it perform well. I therefore conclude that dynami