Case-Injection Improves Response Time for a Real-Time Strategy Game

We present a case-injected genetic algorithm player for Strike Ops, a real-time strategy game. Such strategy games are fundamentally resource allocation optimization problems and our previous work showed that genetic algorithms can play such games by solving the underlying resource allocation problem. This paper shows how we can learn to better respond to opponent actions (moves) by using case-injected genetic algorithms. Case-injected genetic algorithms were designed to learn to improve performance in solving sequences of similar problems and thus provide a good fit for responding to opponent actions in Strike Ops which result in a sequence of similar resource allocation problems. Our results show that a case-injected genetic algorithm player learns from previously encountered problems in the sequence to provide better quality solutions in less time for the underlying resource allocation problem thus improving response time by the genetic algorithm player. This improves the responsiveness of the game and the quality of the overall playing experience.