Trap Avoidance in Strategic Computer Game Playing with Case Injected Genetic Algorithms

We use case injected genetic algorithms to learn to competently play computer strategy games. Such games are characterized by player decision in anticipation of opponent moves and imperfect knowledge of game state. Within the broad goal of developing effective and general methods of anticipatory play, this paper investigates anticipation in the context of trap avoidance in an immersive, 3D strike planning game. Case injection allows acquiring player knowledge from experience and incorporating acquired knowledge into future game play. Results show that with an appropriate representation case injection is effective at biasing the genetic algorithm toward producing plans that both avoid traps and carry out the mission effectively.