Estudi i disseny de personatges virtuals amb comportament intel·ligent orientat a objectius i utilitats

Abstract

[en] Nowdays, studies suggest that the commercial success of a videogame is directly related to the quality and robustness of its AI, which is a challenge for game developers. This project focuses on the design and implementation of intelligent behaviours for Non-Player Characters (NPC) in a game. Both techniques choose an action that will be performed by the character. There is a wide range of actions available to characters. Specifically, the techniques that will be studied are: Goal Oriented Behaviour (GOB): this technique chooses an action based on character’s goals. Each goal has a value indicating how satisfied it is, and performing an action can increase or decrease this value for one or more goals. The chosen action will be the one that affects more positively to the world, leaving the major possible number of goals satisfied. Utility System: this technique choses an action based on utility values of each action. Each action has associated game state variables. For example, the action Go to eat can have associated the variable character’s hunger. The utility of an action depends on the value of the world variables it has associated. This work presents results by means of simulations and an educational game called Fracsland, previously developed in another final project. Initially, Fracsland did not incorporate AI in NPCs, with this project they will be capable of making their own decisions. The chosen NPC to apply the studied techniques is called Lord Barus, the governor of the island. The main idea is to give him an AI to assign quests to the player based on the status of the island. His goals are: keep the population safe, maintain the buildings in decent condition, have the farm animals healthy and possess stocks of food and materials.