Generation of Emotional Behavior for Non-Player Characters - Development of EmoBot for Quake II

Abstract

Recent technology advances in computer game industries demand a development of high-level artificial intelligence. This trend motivates the International Game Developers’ Association to develop a special interest group (SIG) in artificial intelligence (IGDA 2003) and the roundtable of ‘AI in Computer Games’ at the Game Developers Conference (GDC 2003). According to the IGDA SIG, high-level artificial intelligence is defined by the realization of more attractive humanistic features within computer games, such as believable and interesting nonplayer characters (NPCs) that perform complex reasoning and learning in order to exhibit emotions. Similarly, the AAAI 2000 Spring Symposium on AI and Interactive Entertainment discussed issues concerning the generation of human-like characteristics for computer games and toys. Reflecting industrial demands and being inspired by contributions from AAAI symposiums, a set of intelligent agents named EmoBot was developed to generate emotional behaviors for NPCs (Laird 2000; Wilson 2000). EmoBot is embedded within an open-source first-person shooter computer game, ID Software’s Quake II. The following provides a short overview of the AI engine in Quake II (Hooley 2003): Quake II by ID Software uses a simplistic statemachine approach to its artificial intelligence. ... An objective can be interpreted as a “state.” Bots start out with their objective as the world, which signifies the bot will follow pre-determined actions such as guarding or patrolling. Once the bot is aware of a player, the bot’s objective becomes that player.