Understanding Everything NPCs Can Do

Metrics for Action Similarity in Non-Player Characters

Conference paper (2018)

Authors

J.T. Balint Computer Graphics and Visualisation - EEMCS
Jan M. Allbeck George Mason University
Rafael Bidarra Computer Graphics and Visualisation - EEMCS
Research Group
Computer Graphics and Visualisation (EEMCS) (TU Delft)
Copyright: © 2018 J.T. Balint, Jan M. Allbeck, Rafael Bidarra
DOI
help_outline
https://doi.org/10.1145/3235765.3235776
Reasoning Artificial Intelligence in Games Behavior Organization
More Info
expand_more

Published Date

2018

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Faculty

Electrical Engineering, Mathematics and Computer Science

Department

Intelligent Systems

Research Group

Computer Graphics and Visualisation

Abstract

Non-Player Characters (NPCs) have actions that allow them to reason about what they can do in a game and how they can do it. The background information about what they can do are the components of the action, and how they can do it is the form or shape of an action, which may be built up from several sub- actions. The components and shape of an action must be fully defined in a game, which can be tedious when several similar actions are needed. Furthermore, as the number of nuanced actions grows, more pressure is placed on an already constrained reasoning system. By discovering the similarities between actions an NPC can do, a given action set can be intelligently organized with similar components being generalized using an action taxonomy. To understand the similarity between actions we have developed measures of action similarity based on their constituent components and form. From this, we
discover a metric to determine the generalization ability of an organization strategy for an NPC action set. We examine the use of our measures on a previously developed action set to show the nuances between those actions. Lastly, we find that intelligently organizing actions has a positive effect on virtual character reasoning abilities.