Post-graduate theses
Current Record: 21 of 833
|
Identifier |
000461061 |
Title |
Autonomia: a knowledge-based framework for realistic agent behaviours in dynamic video game environments |
Alternative Title |
Autonomia: ένα προγραμματιστικό πλαίσιο που βασίζεται στην γνώση για ρεαλιστικές συμπεριφορές πρακτόρων σε δυναμικά περιβάλλοντα βιντεοπαιχνιδιών |
Author
|
Περβολαράκης, Ζαχαρίας Ε.
|
Thesis advisor
|
Στεφανίδης, Κωνσταντίνος
|
Reviewer
|
Μαγκούτης, Κωνσταντίνος
Ζαμπούλης, Ξενοφών
|
Abstract |
Video games are a popular form of entertainment that offer interactive and
immersive experiences to the players. A key element of these experiences is the
presence of non-player characters (NPCs), which are autonomous agents that
populate the game world and interact with the player and the environment. NPCs
can enhance the realism and diversity of the game scenarios by exhibiting human-
like behaviours that are consistent, adaptive and believable. However, creating
such behaviours is a complex and challenging task that requires a combination of
artificial intelligence (AI) techniques and game design principles. Current methods
and frameworks for NPC decision-making often rely on predefined scripts or rules
that limit the NPC’s capability to adapt to dynamic situations. Moreover, NPCs
usually lack autonomy, as they are unable to pursue their own goals, as well as to
interact with other NPCs or the player. Therefore, there is a need for novel
approaches that can improve the credibility and adaptability of NPC behaviours in
video games.
This work introduces Autonomia, an innovative knowledge-based framework for
realistic agent behaviours in dynamic video game environments. Autonomia is
deeply rooted in the Theory of Mind (ToM), leveraging a knowledge graph to depict
the world's state, with each NPC possessing a replica of this world state in its
“memory”. This “memory” is designed to support higher orders of ToM while
constantly evolving as the NPC perceives the world around it and interprets events.
Autonomia uses a modular system to define the functionality and behaviour of
different types of nodes in the graph, such as physical objects, animals or people.
The framework as structured, allows NPCs to dynamically react to changes in the
environment purely based on its ability to perceive and hold memory. In this
context, Autonomia introduces a new way to model behaviours and goals, enabling
them to be treated as knowledge that can be communed, discovered or even
forgotten just like any other part of the NPC's “memory”. Basing everything on their
acquired knowledge, NPCs utilize a Goal-Oriented Action Planning (GOAP)
algorithm to come up with plans in any dynamic environment.
Lastly, an implementation of Autonomia is provided for the Unity game engine,
including the “Prometheus Tavern” case study, on which a two-part expert-based
evaluation was conducted. The first part confirmed that the provided features and
the architecture of the Autonomia framework deliver solutions that can improve the
credibility of NPC behaviours, whereas the second showed that the agents of the
system have the capability to adapt to their environment and behaviour in a realistic
manner.
|
Language |
English |
Subject |
Autonomous Agents |
|
Game AI |
|
NPC |
|
NPC Behaviours |
Issue date |
2023-12-01 |
Collection
|
School/Department--School of Sciences and Engineering--Department of Computer Science--Post-graduate theses
|
|
Type of Work--Post-graduate theses
|
Permanent Link |
https://elocus.lib.uoc.gr//dlib/3/3/2/metadata-dlib-1701959625-154345-17277.tkl
|
Views |
1026 |