Evolutionary algorithms in particular and computational intelligence techniques in general have been often used to generate both the contents of a game (levels, maps, audiovisual elements, etc.) and the artificial intelligence governing the behavior of non-player characters (NPCs). These two tasks are often approached separately, i.e., the contents are usually generated either independently of the game AI or assuming some fixed AI in order to optimize some property of the game; similarly so, the AI is commonly optimized to work on a certain scenario or collection of scenarios. It is however clear that these two aspects are interrelated in general. For example, the structure of a certain game map may influence the way the AI behaves; furthermore, an evolved AI may be overadapted to particular features of a map. Thus, if the map is optimized to promote some feature of the game, its success will depend on the AIs used and, conversely, an AI optimized to exhibit some interesting behavior will or will not do so depending on the map on which it is deployed.
A typical evolutionary solution to deal with this kind of entangled fitness definitions is the use of competitive co-evolution. By establishing an arms race between AIs and game contents, it is possible to exhert a continuous evolutionary pressure for adapting and discovering new scenarios and strategies. Hence, it can be seen as a holistic self-learning approach. We have precisely followed this idea in a recent paper in which we co-evolve the map layout and the AI behavior for Planet Wars, a RTS game. This paper is entitled
- Competitive Algorithms for Co-evolving both Game Content and AI. A Case Study: Planet Wars (M. Nogueira-Collazo, C. Cotta, A.J. Fernández-Leiva)
and has been published in the IEEE Transactions on Computational Intelligence and AI in Games. Its abstract is as follows:
The classical approach of Competitive Coevolution (CC) applied in games tries to exploit an arms race between coevolving populations that belong to the same species (or at least to the same biotic niche), namely strategies, rules, tracks for racing, or any other. This paper proposes the co-evolution of entities belonging to different realms (namely biotic and abiotic) via a competitive approach. More precisely, we aim to coevolutionarily optimize both virtual players and game content. From a general perspective, our proposal can be viewed as a method of procedural content generation combined with a technique for generating game Artificial Intelligence (AI). This approach can not only help game designers in game creation but also generate content personalized to both specific players’ profiles and game designer’s objectives (e.g., create content that favors novice players over skillful players). As a case study we use Planet Wars, the Real Time Strategy (RTS) game associated with the 2010 Google AI Challenge contest, and demonstrate (via an empirical study) the validity of our approach.
The reference of the paper is:
M. Nogueira-Collazo, C. Cotta, A.J. Fernández-Leiva, Competitive Algorithms for Co-evolving both Game Content and AI. A Case Study: Planet Wars, IEEE Transactions on Computational Intelligence and AI in Games, 2015, online first, DOI: 10.1109/TCIAIG.2015.2499281
Please, check also some of our previous works on Planet Wars.