Using CIGAR for finding effective group behaviors in RTS game
We investigate using case-injected genetic algorithms to quickly generate high quality unit micro-management in real-time strategy game skirmishes. Good group positioning and movement, which are part of unit micro-management, can help win skirmishes against equal numbers and types of opponent units or win even when outnumbered. In this paper, we use influence maps to generate group positioning and potential fields to guide unit movement and compare the performance of case-injected genetic algorithms, genetic algorithms, and two types of hill-climbing search in finding good unit behaviors for defeating the default Starcraft Brood Wars AI. Early results showed that our hill-climbers were quick but unreliable while the genetic algorithm was slow but reliably found quality solutions a hundred percent of the time. Case-injected genetic algorithms, on the other hand were designed to learn from experience to increase problem solving performance on similar problems. Preliminary results with case-injected genetic algorithms indicate that they find high quality results as reliable as genetic algorithms but up to twice as quickly on related maps.
Liu, Siming, Sushil J. Louis, and Monica Nicolescu. "Using CIGAR for finding effective group behaviors in RTS game." In 2013 IEEE Conference on Computational Inteligence in Games (CIG), pp. 1-8. IEEE, 2013.
IEEE Conference on Computatonal Intelligence and Games, CIG