Grupo de Pesquisa em Jogos

Resumo

Overlord é um Orquestrador de Conteúdos, isto é, um sistema que inicia e coordena dois ou mais geradores procedurais de conteúdo. Atualmente, Overlord gerencia geradores de Níveis (dungeons, salas e posicionamento de objetos), Regras (parâmetros de inimigos, como vida, velocidade, dano, tipo de arma e de movimento) e Narrativa (quests). O orquestrador também considera perfis de jogador, e adapta os conteúdos gerados de acordo com o perfil definido.

Publicações

Revista

04/07/2025 • Leonardo T. Pereira, T. Yuji Teoi, Claudio F. M. Toledo
Evaluating a Procedural Content Orchestrator Gameplay Data and Classifying User Profiles

Palavras-chave: procedural content generation, player profiling, content orchestration, gameplay metrics evaluation, machine learning
Abstract • BibTeX • Full Paper

Abstract
Background: Content Orchestration is a novel research field focused on coordinating distinct types of algorithmically generated game content. Purpose: Thus, the lack of research in this area hinders the analysis of gameplay data and player profiling in games with orchestrated content. Methods: This paper is an extension of a work that collected and analyzed gameplay logs of 15 players who played 119 game sections of 12 different dungeons of a top-down action game. The game’s Levels, Rules, and Narrative content were orchestrated and adapted to player profiles defined from a pre-test questionnaire. PCA and clustering techniques were used to highlight relevant gameplay metrics for distinguishing play styles. In this extension, we used the gameplay data alone to train classifiers with and without data augmentation to predict a user’s profile, measuring the accuracy, precision, recall and f1-score with a train-test split and a 5-fold cross-validation for a more robust accuracy. We also implemented data augmentation on our gameplay metrics sample. Results: We identified, through the previous work, two components of PCA explaining a total of 65% of data variability, containing data such as Lock Usage Rate, Enemy Kill Rate, Map Completion, and Completed Immersion Quests. We also found game difficulty as an important level component for impact clustering. Through data augmentation, we achieved novel results, such as a mean accuracy of almost 95%, measured with a 5-fold cross-validation, for the Histogram-based Gradient Boosting classifier when predicting a player’s profile based on their gameplay data, even with our small sample size. Conclusion: Our work guides developers and researchers to choose relevant gameplay metrics to determine players’ play styles. Our extended results suggest that we can predict player’s profiles through gameplay metrics and data augmentation, even for small samples. More studies are needed to validate our findings, with a larger and more diverse player-base.
BibTeX • Copiar
```
@article{pereira:2025:jis, 
place={Porto Alegre, RS}, 
title={Evaluating a Procedural Content Orchestrator Gameplay Data and Classifying User Profiles}, 
volume={16}, 
url={https://journals-sol.sbc.org.br/index.php/jis/article/view/5630}, 
DOI={10.5753/jis.2025.5630}, 
number={1}, 
journal={Journal on Interactive Systems}, 
author={Pereira, Leonardo Tórtoro and Teoi, T. Yuji and Toledo, Claudio Fabiano Motta}, 
year={2025}, 
month={Jul.}, 
pages={490–504} 
}
```
12/10/2022 • Leonardo T. Pereira, Breno M. F. Viana, Claudio F. M. Toledo
A System for Orchestrating Multiple Procedurally Generated Content for Different Player Profiles

Palavras-chave: software architecture, procedural content generation, Content orchestration, games, player classification
Abstract • BibTeX • Full Paper

Abstract
We propose a system able to orchestrate the procedural generation of three creative facets (levels, narrative, and rules), using quality-diversity algorithms to promote diversity while converging to the desired solutions and able to satisfy different player's profiles. We aim to facilitate the project's reproducibility and extension by presenting the system's architecture and high-level design decisions and showcasing how we decoupled it enough to change algorithms. Our generation starts with a grammar-based quest generator that works as a frame for two quality-diversity algorithms: a level generator, which also creates locked-door puzzles, and an enemy generator, which selects their visuals based on their weapons. A rule-based profile analysis uses a pretest questionnaire to classify the player into four profiles, and the quest generator creates different content for each of them. We use a game prototype to test our system with over 83 players, playing over 204 levels and answering a posttest questionnaire after each. Results show the game and contents were well received, and there was a difference in perception between profiles. Players who played content based on their profile enjoyed it more than those playing content from another profile.
BibTeX • Copiar
```
@ARTICLE{pereira:2022:tog,
author={Pereira, Leonardo Tortoro and Viana, Breno M. F. and Toledo, Claudio F. M.},
journal={IEEE Transactions on Games}, 
title={A System for Orchestrating Multiple Procedurally Generated Content for Different Player Profiles}, 
year={2024},
volume={16},
number={1},
pages={64-74},
keywords={Video games;Visualization;Grammar;Software architecture;
Procedural generation;Content orchestration;games;player classification;
procedural content generation},
doi={10.1109/TG.2022.3213781}}
```
15/10/2021 • Leonardo T. Pereira, Paulo V. S. Prado, Rafael M. Lopes, Claudio F. M. Toledo
Procedural generation of dungeons’ maps and locked-door missions through an evolutionary algorithm validated with players

Palavras-chave: Evolutionary algorithm, Procedural content generation, Level generation, Mission generation, Adaptive game content
Abstract • BibTeX • Full Paper

Abstract
The present research introduces an evolutionary algorithm able to procedurally generate dungeon maps containing locked door missions. The evolutionary algorithm evolves a tree structure, which encodes dungeons, aiming to generate levels close to the input configuration provided by a game designer. The tree structure holds information about the number of rooms, connections between them, and their position within a 2D map. The proposed encoding also allows evolving semantic information about the narrative of the game. This is done by feasibly setting keys and locks throughout the dungeons for locked door missions. The generated dungeons are then evaluated computationally and as a proof of concept using an adventure game prototype. A total of 70 players evaluated the contents and the results show that the procedurally generated levels are perceived as more human-made, fun, and difficult than their human-made counterparts for most cases.
BibTeX • Copiar
```
@article{pereira:2021,
title = {Procedural generation of dungeons’ maps and 
locked-door missions through an evolutionary algorithm validated with players},
journal = {Expert Systems with Applications},
volume = {180},
pages = {115009},
year = {2021},
issn = {0957-4174},
doi = {https://doi.org/10.1016/j.eswa.2021.115009},
url = {https://www.sciencedirect.com/science/article/pii/S0957417421004504},
author = {Leonardo Tortoro Pereira and Paulo Victor de Souza Prado and 
Rafael Miranda Lopes and Claudio Fabiano Motta Toledo},
keywords = {Evolutionary algorithm, Procedural content generation, 
Level generation, Mission generation, Adaptive game content},
}
```

Conferência

2024

07/10/2024 • Leonardo T. Pereira, T. Yuji Teoi, Claudio F. M. Toledo
Towards Evaluating a Procedural Content Orchestrator Gameplay Data to Differentiate User Profiles

Palavras-chave: procedural content generation, player profiling, gameplay metrics evaluation
Abstract • BibTeX • Full Paper

Abstract
We tested a procedural content orchestration algorithm against 15 anonymous users, against 12 different dungeons, played 119 times in total. We used questionnaires to collect data regarding player profiles, and gameplay data to analyze if could identify profiles using them only. Using PCA and clustering techniques, we were able to identify the most important attributes one may collect from gameplay data to analyze and differentiate play-styles. We also identified that the dungeon's characteristics have a heavy influence on analyzing profiles through gameplay, and a more controlled environment may be needed to identify player profiles. More data and further analysis are needed to extract player profiles from gameplay data, but preliminary results show promise.
BibTeX • Copiar
```
@inproceedings{pereira:2024:wiplay,
author = {Leonardo Pereira and Yuji Teoi and Claudio Toledo},
title = { Towards Evaluating a Procedural Content Orchestrator Gameplay Data to Differentiate User Profiles},
booktitle = {Anais do III Workshop sobre Interação e Pesquisa de Usuários no Desenvolvimento de Jogos},
location = {Brasília/DF},
year = {2024},
keywords = {},
issn = {0000-0000},
pages = {75--86},
publisher = {SBC},
address = {Porto Alegre, RS, Brasil},
doi = {10.5753/wiplay.2024.245483},
url = {https://sol.sbc.org.br/index.php/wiplay/article/view/31181}
}
```
30/09/2024 • T. Yuji Teoi, Leonardo T. Pereira, Claudio F. M. Toledo
Towards Adapting a Content Orchestrator to a Different GameGenre: Generating Levels, Rules, and Narrative for Diverse Player Profiles from a Top-down Adventure to a 2D Platformer

Palavras-chave: procedural content generation, content orchestration, digital games, platformer
Abstract • BibTeX • Full Paper

Abstract
Procedural Content Generation for multiple game content facets is a challenge for the game industry and academia. A content orchestrator is a software that can manage different procedural content generators, mixing their outputs while maintaining coherence and feasibility. We adapted a content orchestrator, originally meant for a top-down adventure game, to a 2D platformer. Both versions procedurally generate Levels, Rules, and Narrative, while adapting to distinct player profiles. A pre-test questionnaire is used to evaluate the player profile, and a post-test questionnaire is used to evaluate the game prototype we developed for this experimental purpose and the procedurally generated content. Results show the game was fun, challenging, interesting to explore, and with a moderate difficulty. Although with a limited sample, our results indicate the system was able to target content based on profiles. Therefore, this is a first step into understanding how a content orchestrator can be adapted to different game genres.
BibTeX • Copiar
```
@inproceedings{teoi:2024,
author = {Yuji Teoi and Leonardo Pereira and Claudio Toledo},
title = { Towards Adapting a Content Orchestrator to a Different GameGenre: 
Generating Levels, Rules, and Narrative for Diverse Player Profiles from a 
Top-down Adventure to a 2D Platformer},
booktitle = {Anais do XXIII Simpósio Brasileiro de Jogos e Entretenimento Digital},
location = {Manaus/AM},
year = {2024},
keywords = {},
issn = {0000-0000},
pages = {396--415},
publisher = {SBC},
address = {Porto Alegre, RS, Brasil},
doi = {10.5753/sbgames.2024.240282},
url = {https://sol.sbc.org.br/index.php/sbgames/article/view/32327}
}
```
19/02/2022 • Breno M. F. Viana, Leonardo T. Pereira, Claudio F. M. Toledo
Illuminating the Space of Enemies Through MAP-Elites

Palavras-chave: enemy generation, procedural content generation, video game, evolutionary algorithm, map-elites
Abstract • BibTeX • Preprint

Abstract
Action-Adventure games have several challenges to overcome, where the most common are enemies. The enemies' goal is to hinder the players' progression by taking life points, and the way they hinder this progress is distinct for different kinds of enemies. In this context, this paper introduces an extended version of an evolutionary approach for procedurally generating enemies that target the enemy's difficulty as the goal. Our approach advances the enemy generation research by incorporating a MAP-Elites population to generate diverse enemies without losing quality. The computational experiment showed the method converged most enemies in the MAP-Elites in less than a second for most cases. Besides, we experimented with players who played an Action-Adventure game prototype with enemies we generated. This experiment showed that the players enjoyed most levels they played, and we successfully created enemies perceived as easy, medium, or hard to face.
BibTeX • Copiar
```
@article{viana2022illuminatingenemies,
  title={Illuminating the Space of Enemies Through MAP-Elites},
  author={Viana, Breno M. F. and Pereira, Leonardo T. and Toledo, Claudio F. M.},
  journal={arXiv preprint arXiv:2202.09615},
  year={2022}
}
```
18/02/2022 • Breno M. F. Viana, Leonardo T. Pereira, Claudio F. M. Toledo
Illuminating the Space of Dungeon Maps, Locked-door Missions and Enemy Placement Through MAP-Elites

Palavras-chave: evolutionary algorithm, map-elites, procedural content generation, level generation, mission generation, video game
Abstract • BibTeX • Preprint

Abstract
Procedural Content Generation (PCG) methods are valuable tools to speed up the game development process. Moreover, PCG may also present in games as features, such as the procedural dungeon generation (PDG) in Moonlighter (Digital Sun, 2018). This paper introduces an extended version of an evolutionary dungeon generator by incorporating a MAP-Elites population. Our dungeon levels are discretized with rooms that may have locked-door missions and enemies within them. We encoded the dungeons through a tree structure to ensure the feasibility of missions. We performed computational and user feedback experiments to evaluate our PDG approach. They show that our approach accurately converges almost the whole MAP-Elite population for most executions. Finally, players' feedback indicates that they enjoyed the generated levels, and they could not indicate an algorithm as a level generator.
BibTeX • Copiar
```
@article{viana2022illuminatinglevels,
  title={Illuminating the Space of Dungeon Maps, Locked-door Missions and Enemy Placement Through MAP-Elites},
  author={Viana, Breno M. F. and Pereira, Leonardo T. and Toledo, Claudio F. M.},
  journal={arXiv preprint arXiv:2202.09301},
  year={2022}
}
```
18/10/2021 • Leonardo T. Pereira, Breno M. F. Viana, Claudio F. M. Toledo
Procedural Enemy Generation through Parallel Evolutionary Algorithm

Palavras-chave: enemy generation, procedural content generation, video games, parallel evolutionary algorithm
Abstract • BibTeX • Full Paper • PDF

Abstract
This research presents a Parallel Evolutionary Algorithm (PEA) that generates enemies with diverse characteristics, such as the enemy's health, weapons, and movement. Our PEA aims to create enemies matching their difficulty degrees with the difficulty goal given as input parameter. We designed our algorithm in this way to be future used in an online adaptive generation system. We experimented with a set of generated enemies with an Action-Adventure game prototype as a testbed. The results show that players evaluated our approach positively, successfully creating enemies considered easy, medium, or hard to face, as defined by their original fitness' target value. Besides, the players found the game fun to play for all difficulty levels played, and the perceived challenge rose as the PEA fitness was higher. In terms of performance results, our PEA converged into the input solution in less than a second for most cases, denoting its future use in online adaptive applications.
BibTeX • Copiar
```
@INPROCEEDINGS{9637547,
  author={Pereira, Leonardo T. and Viana, Breno M. F. and Toledo, Claudio F. M.},
  booktitle={2021 20th Brazilian Symposium on Computer Games and Digital Entertainment (SBGames)},
  title={Procedural Enemy Generation through Parallel Evolutionary Algorithm},
  year={2021},
  volume={},
  number={},
  pages={126-135},
  doi={10.1109/SBGames54170.2021.00024}
}
```

Tese

2022

12/10/2022 • Leonardo T. Pereira, Breno M. F. Viana, Claudio F. M. Toledo
Arquitetura e desenvolvimento de um sistema de geração procedural de múltiplos conteúdos para jogos eletrônicos em tempo real

Palavras-chave: Algoritmo evolutivo, Arquitetura de software, Design de sistemas, Geração de conteúdo em tempo real, Geração procedural de conteúdo
Abstract • BibTeX • Full Paper

Abstract
É apresentado um sistema capaz de orquestrar a geração procedural de múltiplos conteúdos, usando diferentes técnicas de criatividade computacional, e capaz de gerar os conteúdos em tempo real. Os conteúdos gerados são calabouços, o posicionamento de inimigos, chaves e fechaduras em suas salas, os inimigos que serão colocados no calabouço, missões para o jogador completar, e as salas do calabouço. São usados algoritmos evolutivos, MAP-Elites, Gramáticas Formais, Cadeias de Markov e Autômatos Celulares para criar tais conteúdos. Eles são organizados por um orquestrador e colocados em um protótipo de jogo. No último estágio do sistema, o conteúdo pode ser gerado em tempo real, tão logo quanto são coletados os dados do jogador para usar de entrada, através de uma análise de perfis baseada em regras. Também é descrita uma arquitetura e design de sistema, que permitiram a adição de novos geradores e a troca de algoritmos para cada gerador com certa facilidade. Diferentes experimentos com jogadores foram realizados para testar diferentes algoritmos e configurações do sistema, todos com resultados positivos, e com os jogadores divertindo-se ao jogar os conteúdos resultantes. Jogadores que jogaram o conteúdo baseado em seus perfis divertiram-se mais do que aqueles que jogaram conteúdo de outro perfil. Isso mostra que nosso identificador de perfis com base em regras pode guiar os geradores para bons resultados.
BibTeX • Copiar
```
@phdthesis{pereira:2022:phd,
  author    = {Leonardo Tortoro Pereira},
  title     = {Arquitetura e desenvolvimento de um sistema de geração procedural 
  de múltiplos conteúdos para jogos eletrônicos em tempo real},
  school    = {Universidade de São Paulo},
  address   = {São Carlos},
  year      = {2022},
  month     = dec,
  doi       = {https://doi.org/10.11606/T.55.2022.tde-28032023-091310},
  url       = {https://doi.org/10.11606/T.55.2022.tde-28032023-091310},
  language  = {pt-BR},
  note      = {Tese de Doutorado, Instituto de Ciências Matemáticas e de Computação.},
  keywords  = {Algoritmo evolutivo, Arquitetura de software, Design de sistemas, 
  Geração de conteúdo em tempo real, Geração procedural de conteúdo},
  advisor   = {Toledo, Cláudio Fabiano Motta},
  committee = {
    Toledo, Cláudio Fabiano Motta (Presidente),
    Liapis, Antonios,
    Ontanon, Santiago,
    Osório, Fernando Santos
  }
}
```

Equipe

Leonardo T. Pereira

Breno M. F. Viana