The use of Fuzzy Logic for Artificial Intelligence in Games

Game artificial intelligence (game AI) is the branch of videogame development that is concerned with empowering games with the illusion of intelligence. Game AI borrows many techniques from the broader field of AI, from simple finite state machines to state-of-the-art evolutionary algorithms. Among these techniques, fuzzy logic is one of the tools that must be present in the arsenal of a good videogame AI developer, due to the simplicity of its formulation coupled with its expressive power. In this paper, we introduce the field of game AI and review the use of fuzzy logic in games, looking both at industry and research. We outline its benefits, address its shortcomings and present its practical uses. 1 Artificial Intelligence and Games The field of game artificial intelligence encompasses all the techniques and methods for injecting intelligence into video games. To game AI pertain many different aspects of a videogame: animation control, steering, flocking, pathfinding, planning, procedural generation, tactical and strategic thinking, and learning [21, 6]. All these aspects share the same basis, they pose problems whose efficient solution requires AI algorithms. The purpose of game AI is to bring life to the non-playable characters (NPCs) present in the games, such as the enemy troops in a strategy game or the merchants in a peaceful AI-controlled village. Game AI is just one of the branches of the broader field of artificial intelligence. Academic AI research always suffered from a difficulty to define artificial intelligence itself, although it can be typically identified in the study and recreation of human-like or human-level cognitive processes and in the capability of these processes to learn [16]. According to Alan Turing, who is considered the father of artificial intelligence, an agent is intelligent if its behavior cannot be distinguished from that of a human [33]. Game AI adds another piece to the puzzle of defining AI, as it is not concerned with the actual existence of intelligence, but just with the illusion of it [29]. For a game to be successful, we do not really need highlyintelligent, human-like, possibly unbeatable expert opponents, what is required is a compelling adversary, an AI agent that is fun to play against and that is at least not obviously stupid. Game AI, like all other aspects of a videogame, exists to propel the higher goal of a videogame: to be fun and entertain. In other words, the motto of the game AI society is often thought to be If the player cannot see it, why do it? [21]. The context is however not so simple nowadays, as partly due to successful products that have brought forward advanced AI, partly due to the ubiquitous online multi-player options that have changed the habits of players, players are now more demanding in regards to AI and more and more game developers have turned their attention from scripted and predictable agents to more human-like agents, capable of learning and thus, in the classical AI definition, intelligent. Another peculiarity of game AI lies in the fact that games are, for the most part, subject to real-time constraints and thus there is little space for slow, offline techniques. This fact shares similitudes to the AI used in robotics and control systems [25], and this is the reason why the same techniques, in one form or another, can be found in all these fields. At last, we must remember that video games are part of the entertainment industry, an industry well-known for its fast advancements, sudden turns of trends and tight development schedules. This leaves little room for highly innovative yet not throughly tested AI techniques that may hide nasty surprises and detour development. Due to these differences, the game AI field has been advancing in a different but parallel direction in respect to academic AI, borrowing ideas and methods from the latter while always keeping a practical approach in mind [16, 32]. In this paper, we discuss the application of a specific technique to videogame AI: fuzzy logic. In the next section, we examine the current state of the game AI field in order to provide a context for the discussion on fuzzy logic. We introduce the concepts of fuzzy logic and detail its applications for game AI in section 3. In section 5, existing uses of fuzzy logic in the industry and in research are reported and analyzed. In section 6, we draw conclusions on the current situation about the use of fuzzy logic in games. 2 The current state of Game AI Considered as a last-minute addition in the eighties and nineties, game AI has instead been gaining attention from both academics and industry during the last decade. Traditionally, games have been pushing on their visual and audial compo-