Antonios Liapis: Projects

Computer-Aided Design

Sentient Sketchbook

Sentient Sketchbook is a tool which supports the design of map sketches via an intuitive interface, via real-time feedback regarding the map's playability or balance, and via the suggestion of alternative map designs.

Computer-Aided Design

Sentient World

Sentient World is a tool which allows a designer to progressively add details to a rough sketch through the process of iterative refining.

Divergent Search

DeLeNoX

DeLeNoX applies divergent evolutionary search to a generative algorithm, using a notion of novelty based on properties of the content that the generator has produced so far.

Knowledge Engineering

Data Adventures

The Data Adventures trilogy aims to generate adventure games based on open data (for instance Wikipedia articles, Wikimedia commons and OpenStreetMap).

Computer-Aided Design

LLMaker

LLMaker is a chat-only tool facilitating a human user to design games using natural language. LLMaker is powered by Large Language Models and Large Media Models.

Procedural Content Generation

CrawLLM

CrawLLM is a dungeon crawler game with card-based combat mechanics, where Large Language Models and Large Media Models are used to provide novel themes for the narrative and graphics.

Procedural Content Generation

Sonancia

Sonancia is a system built for generating multiple facets of horror games, with the intention of creating tense and frightful experiences.

Procedural Content Generation

Surrogate-based Generation of Arenas (SuGAr)

SuGAr aims to automatically generate or adapt elements of a first-person shooter (FPS) game based on learned mappings between level structures, game parameters, and gameplay outcomes.

Computer-Aided Design

SuSketch

SuSketch is a game and level design tool that uses machine learning to predict gameplay outcomes and to generate suggestions for the designer in a simple Firest Person Shooter game.

User Modeling

AffRooms Dataset

The AffRooms dataset collects time-continuous affect annotations and matches the changes in affect between rooms with the design features of each room (e.g. lighting, curvature).

Agent Control

Dungeons & Replicants

This project trained 231 models imitating individual players, and uses them as artificial playtesters in order to (a) find the ideal challenge level for an AI enemy and (b) to improve the balance between character classes.

User Modeling

Pixels 2 Affect

In this line of research, the gameplay footage (screen pixels, and optionally audio) is used as the sole input to deep networks which predict the arousal of the player.

Agent Control

Go-Blend

Go-Blend is a reinforcement learning algorithm able to generate agents that exhibit various behavioral and affective patterns as specified by its rewards functions.

User Modeling

Arousal Game Annotation (AGAIN) Dataset

The AGAIN dataset contains over 1000 annotated gameplay sessions, where players themselves annotate their playthrough in terms of the arousal they felt at each point of the game.

User Modeling

Predicting Player Motivation in Tom Clancy's The Division

In a collaboration with Ubisoft, gameplay data from players of Tom Clancy's The Division are used to train models that predict a player's self-reported levels of competence, autonomy, relatedness and presence.

'Serious' Games

ArtBot

ArtBot is an educational game intended to improve AI literacy through a series of tasks requiring the application of supervised and reinforcement learning.

'Serious' Games

Iconoscope

Iconoscope is a multi-player digital game intended to infuse creativity to the learning process, played locally in a group or as a standalone web app.

Divergent Search

Soft Robot Evolution with Surprise Search

This project tests the robustness and efficiency of divergent search algorithm such as MAP-Elites, novelty search, surprise search, and their combination in a challenging testbed.

User Modeling

MiniDungeons

MiniDungeons is a simple turn-based roguelike puzzle game, implemented as a benchmark problem for modeling decision-making styles of human players.

User Modeling

MiniDungeons 2

MiniDungeons 2 is a turn-based roguelike puzzle game, designed to provide a complex yet fully observable and deterministic environment for modeling decision-making styles of human players.

User Modeling

Automated Spaceship Design

The Spaceship Design tool is an experiment in interactive evolution where the user's visual taste is captured via adaptive computational models.

Divergent Search

Weapons Generated for Surprise

Balanced and efficient sets of weapons for the Unreal Tournament III shooter game are generated through a constrained divergent search algorithm driven by surprise.

User Modeling

Mazing

This project explores the player's emotional theory of mind regarding the frustration level of an artificial agent acting as an opponent in a prototype survival game.

Procedural Content Generation

AudioInSpace

AudioInSpace is a space shooter game where the game's soundtrack affects the behavior of the player's weapons and the weapon particles and the player's firing patterns in turn affect the musical notes played.

Computer-Aided Design

Coralize

Coralize is a plugin for the Unity 3D game engine, and encompasses a library of generators for marine organisms, i.e. stony corals, soft corals and sponges.

Procedural Content Generation

Multi-floor FPS Level Generation

This project leverages genetic algorithms to evolve a shooter level's architecture and the placement of powerups and player spawnpoints, generating levels with one floor or two floors.