Return to Dr. Jacob Schrum's Homepage

Generating Lode Runner Levels With Generative Adversarial Networks (GANs)

This page presents research in Procedural Content Generation via Machine Learning done by undergraduate student Kirby Steckel as part of Southwestern University's Summer research program SCOPE. The research makes use of the MM-NEAT software package, though neural networks are not evolved. Rather, real-valued latent vectors are evolved and sent as input to pre-trained GANs which output Lode Runner levels. Lode Runner is a classic platform game where the player must run around and climb ladders to collect treasure items, while avoiding enemy agents. All GANs used in this study were trained on data from the Video Game Level Corpus. The focus of the study is on how the training set affects the quality of the levels discovered by the Quality Diversity Evolutionary Algorithm MAP-Elites (Multidimensional Archive of Phenotypic Elites). The primary focus is on the size of the training set, but one training set also contained a stylistically interesting subset of levels, namely those with obstacles laid out to look like English words. Different approaches are analysed in terms of how many phenotypically diverse levels they produce, and how many of those levels are beatable. Levels are only kept in the MAP-Elites archive if they are unique in terms of several dimensions of variation: percentage of ground tiles, number of treasure chests, and number of enemies.

Presentations

Here is a video presentation that Kirby did for students and faculty at the Math/CS department's weekly lecture series.


Here is the 5-minute presentation submitted for the poster session at the virtual 2021 Genetic and Evolutionary Computation Conference.

On 5 Levels

The GAN used to evolve these levels was only trained on the first 5 levels of the game.

On 20 Levels

The GAN used to evolve these levels was only trained on the first 20 levels of the game.

On 50 Levels

The GAN used to evolve these levels was only trained on the first 50 levels of the game.

On 100 Levels

The GAN used to evolve these levels was only trained on the first 100 levels of the game.

On 150 Levels

The GAN used to evolve these levels was trained on all 150 levels of the game.

Words Present

The GAN used to evolve these levels was trained on the subset of levels where clearly visible words appear in the level.

Associated Publications


Peer-Reviewed Journal Articles


Peer-Reviewed Conference Publications


Extended Abstracts


Technical Reports


Dagstuhl Reports


Undergraduate Poster Presentations Supervised


Associated Movies and Images


Miscellaneous Content

  • Summer 2021: Quality Diversity and Creative Divergent Search: SCOPE Research Presentation made by my SCOPE Summer research students to present to other SCOPE students
  • Fall 2020: Procedural Content Generation for Games with Generative Adversarial Networks: Presentation for the Games AI Research Group at Queen Mary University of London (video)
  • Fall 2020: Interactively Evolving Video Game Levels with Generative Adversarial Networks: 403 Lecture for Math and CS Department
  • Fall 2020: Evolving Mega Man Levels with Generative Adversarial Networks: Virtual SCOPE Open House Website
  • Fall 2020: Evolving Lode Runner Levels with Generative Adversarial Networks: Virtual SCOPE Open House Website
  • Summer 2020: Generating Video Game Levels Using AI: SCOPE Research Presentation made by my SCOPE Summer research students to present to other SCOPE students
  • Summer 2019: Playing and Creating Games With Deep Neural Networks: "Mad Science Monday" presentation made by my SCOPE Summer research students to present to other SCOPE students
  • Fall 2018: Evolutionary Computation Applied to Digital Entertainment and the Arts, poster presented at the President's Appreciation Celebration for Southwestern University donors.
  • Fall 2018: Evolving Mario Levels in the Latent Space of a Deep Convolutional Generative Adversarial Network: presented to Southwestern University students as a 107 Lecture.
  • Summer 2018: The machines have taught themselves to make Mario levels, article in Fast Company about my recent GECCO 2018 paper on generating levels for Super Mario Bros.
  • Spring 2018: Bored with your video game? Artificial intelligence could create new levels on the fly, article in Science about my recent GECCO 2018 paper on generating levels for Super Mario Bros.
  • Spring 2018: Doom and Super Mario could be a lot tougher now AI is building levels, article in The Register about my recent GECCO 2018 paper on generating levels for Super Mario Bros.

  • Last Updated: 9/16/2020