Thesis

Most modern Role-playing video games (RPGs) include an extensive virtual world and interesting narratives that the player can immerse himself in. A captivating and interactive narrative experience is essential in a successful RPG. However, when looking at the options that the player can take to change the virtual environment or the general direction of the narrative, those often feel limited or inconsequential. Furthermore, some content can feel generic and created to please all players, as it opposed to a more personalized experience where the player actions help shape the story. The AAA video game industry have often disregarded this fact and developed very basic Artificial Intelligent and Player Modeling systems that often do not meet the required expectations, despite the fact that these have proven - in academic research - to enhance player interest and expected enjoyment. We do believe that a custom experience, different across player with different traits, motivations and preferences can add more replay value, amusement and better storytelling to any RPG. In this work, we present a Player Modeling architecture that uses a Machine Learning instance that analyses player actions and interactions with the virtual world and associates them with a player profile, in order to create a tailored experience that should provide better enjoyment and immersion for the player. This system was implemented in the popular RPG title The Elder Scrolls V: Skyrim and released as a game modification (mod), which was met with extremely positive feedback by the player community.

In this document, a study about the requirements of a military simulator for officer training is presented. In order to understand what those requirements are, we had interviews with military officers responsible for officer training and analyzed wargames existent in the market, including tabletop games, video games and military simulators in use by military forces. The study results in a comparison between the different military simulators in the market and a set of requirements for a military simulator. We also studied existing commercial frameworks which are low-priced or free, and allow for the development of a constructive simulator. Uniting both analyses, a proposal for a military simulator is presented. The proposed system has the advantage of being more affordable than existing military simulators. To demonstrate the viability of using one of the studied frameworks to develop a military simulator, a prototype was developed and tested. From their results, it can be concluded that the developed model can fulfill the proposed objectives.

When designing synthetic characters that interact with human users, it is becoming more and more important to consider the correct expression of the affective states of these characters, if we want to achieve more believable communication. This work studies the various work done in the field of character animation and interaction, and aims to produce a companion module for a virtual tutoring application, capable of supporting the levels of believability necessary during a dialog between a human and two synthetic virtual characters. We explore the process in which we coordinate both head motion and expression of emotional modalities, as a means to reach our desired goal. We also study the impact of this work by presenting two experiments, that assist in comparing the initial (prior to our changes) and final versions of the application, as well as highlighting the effects our changes introduced on the successful emotional recognition of the emotions of happiness, sadness, anger, fear, surprise and disgust.

Video game development is one of the fastest growing businesses, and appealing to a large audience is one of the key factors to a successful game. Recent research in the field of video game design proposes adapting games to the player's profile, which has the potential to broaden a game's target audience. Companion NPCs are an element of games that, when present, have a large potential influence on the player's experience. This work, takes the Media Equation finding that the law of similarity-attraction applies to relationships between media and people, and applies it to video game NPCs. We propose adapting companion NPC to the player's profile, by aligning the NPCs' assertiveness level to the player's own. We also present a methodology for Media Equation findings to be tested in the context of video games. We developed a testbed game, a 2D puzzle platformer with a companion NPC. We also developed two versions of the NPC's behavior, one for each end of the assertiveness scale. We conducted a 2x2, between-subjects experiment (n=48), in which Assertive and Non-Assertive subjects were randomly matched with one of the NPCs. Subjects recognized the NPC's personality type, giving a significantly higher assertiveness score to the NPC endowed with assertive characteristics. Non-Assertive players reported significantly higher Tension scores when interacting with the Assertive NPC than when interacting with the Non-Assertive NPC. However, based on assertiveness level alignment, there was neither a significant difference in the enjoyment of the experience nor in the player's affinity for the NPC.