Non-player character companions in video games should cooperate with players, understand them, and follow their lead during gameplay. In current games, however, companions tend to exhibit mainly static behaviours, and rarely live up to player expectations. In general, our work is aimed at improving this situation, developing both techniques and tools which allow companion NPCs to behave more appropriately, respecting player preferences and offering a more immersive gameplay for players. Problem Introduction Non-Player Characters (NPC) bring interesting challenges and problems when evolving side by side with human players. For example, in First Person Shooter (FPS) games, a given NPC playing the role of a companion has to behave in a certain way to offer support to the player and help the player during challenging moments. Unfortunately, companion behaviour is often unchanging, static, and frequently very limited as well, resulting in companions that, as neatly summarized by Bakkes, are overly superficial: they may coexist with the player, but they often fail to appropriately cooperate (Bakkes, Spronck, and Postma 2005), reducing their value to players, and interfering with immersion. In this proposal, we are interested in using different Artificial Intelligence (AI) techniques to solve problems caused by NPC companions and to improve the state of game companions. In other words, companions should behave as human players are expecting them to, mimicking human companions behaviour or choosing best response behaviour. To achieve this broad goal, we need both improved AI techniques for companions, appropriate to their supporting role in various situations, as well as sufficient knowledge of the player’s state and goals to select the right companion’s behaviour. This results in a complex design challenge, involving the creation of multiple and varied companion AI components, the need to detect or predict player intention, and for the companion to adapt appropriately. We organize our approach to this problem around the following three goals or themes. Copyright c © 2013, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. • Creating and adapting companion behaviour. This includes defining combat and non-combat behaviours that correlate and change with respect to the player’s in-game intention and behaviours. In this we want to create a companion who is there to improve the player’s immersion and help accomplish harder challenges, while leaving most of the work load to the player. • Player metrics & plan recognition. Knowledge of the player’s intent is crucial to the decision making process of the companion, since the companion should cooperate with the player, and wrong beliefs about the player will negatively influence the player’s experience and immersion. We thus aim to develop player monitoring and prediction techniques that can help the companion act most appropriately. • AI game design tools. The problems we address are complex, and exploration will benefit from tools that can facilitate the research and design processes by showing and quantitatively evaluating different algorithms and game situations. Our approach thus includes development of non-trivial tool support, with the expectation that this will aid research, and also be of value to game designers, as they also need to perform interactive and iterative refinement of game levels and companion behaviours.
[1]
Robert P. Goldman,et al.
A Bayesian Model of Plan Recognition
,
1993,
Artif. Intell..
[2]
Pieter Spronck,et al.
Best-Response Learning of Team Behaviour in Quake III
,
2005
.
[3]
Robin Hunicke,et al.
The case for dynamic difficulty adjustment in games
,
2005,
ACE '05.
[4]
Stefan J. Johansson,et al.
Using multi-agent potential fields in real-time strategy games
,
2008,
AAMAS.
[5]
Michael Buro,et al.
On the Complexity of Two-Player Attrition Games Played on Graphs
,
2010,
AIIDE.
[6]
Rafael Bidarra,et al.
In Press: Ieee Transactions on Computational Intelligence and Ai in Games Adaptivity Challenges in Games and Simulations: a Survey
,
2022
.
[7]
Leslie Pack Kaelbling,et al.
POMCoP: Belief Space Planning for Sidekicks in Cooperative Games
,
2012,
AIIDE.
[8]
Pieter Spronck,et al.
Player behavioural modelling for video games
,
2012,
Entertain. Comput..
[9]
Zoran Popovic,et al.
Automated redesign of local playspace properties
,
2013,
FDG.
[10]
Roger Crawfis,et al.
Optimal cover placement against static enemy positions
,
2013,
FDG.