Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
41 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
41 tokens/sec
o3 Pro
7 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Evolving Flying Machines in Minecraft Using Quality Diversity (2302.00782v2)

Published 1 Feb 2023 in cs.NE and cs.AI

Abstract: Minecraft is a great testbed for human creativity that has inspired the design of various structures and even functioning machines, including flying machines. EvoCraft is an API for programmatically generating structures in Minecraft, but the initial work in this domain was not capable of evolving flying machines. This paper applies fitness-based evolution and quality diversity search in order to evolve flying machines. Although fitness alone can occasionally produce flying machines, thanks in part to a more sophisticated fitness function than was used previously, the quality diversity algorithm MAP-Elites is capable of discovering flying machines much more reliably, at least when an appropriate behavior characterization is used to guide the search for diverse solutions.

Evolving Flying Machines in Minecraft Using Quality Diversity: A Technical Overview

The objective of this paper is to explore the application of evolutionary algorithms to the automatic generation of flying machines within the digital environment of Minecraft. Utilizing Minecraft as a platform for experimentation provides a robust testbed due to its rich rule-based system and creative potential. The research presented leverages both fitness-based evolution and Quality Diversity (QD) algorithms, specifically MAP-Elites, to reliably evolve functional flying machines—a challenge previously unmet by the initial EvoCraft API framework.

Technical Approach

Encoding and Fitness Function: The authors have implemented a genome encoding strategy that relies on real-valued vectors representing potential block placements. Each vector encodes block presence, type, and orientation, directly correlating with blocks placed in a 3D voxel-based space within Minecraft. The fitness function is designed to evaluate movement over time and awards success to machines that demonstrate sustained directional movement—characteristic of a flying machine.

Quality Diversity Search with MAP-Elites: The innovative application of MAP-Elites is instrumental in this exploration. MAP-Elites maintains an archive of diverse solutions by categorizing them into behavioral niches, evaluated through specific characterizations such as block count, negative space, and piston orientation. This approach allows the search to explore a broad range of machine configurations, advancing beyond the limitations of a traditional fitness-based evolutionary paradigm that often converges prematurely.

Experimental Results

The research delineates experimental comparisons between different characterizations of MAP-Elites and a pure fitness-based approach. The results unequivocally demonstrate that the MAP-Elites method using Piston Orientation as a behavioral characterization outperforms other methods. This setup produced flying machines more consistently, even across varied directional capabilities.

Ultimately, the paper reports substantial achievements in this domain:

  • With the observer block set, Piston Orientation MAP-Elites had a 93.33% success rate in evolving flying machines, outperforming pure fitness-based evolution.
  • Directional diversity was not only achieved but was significantly enhanced, with some individual runs generating machines capable of flying in four distinct directions.

The aforementioned results underscore the efficacy of employing a diversity-driven search to explore complex state spaces where direct fitness optimization falls short.

Implications and Future Developments

From a theoretical perspective, this work contributes to the understanding of how Quality Diversity algorithms can be utilized for procedural content generation tasks. It illustrates the significance of incorporating diverse behavioral characterizations in adapting search methods to complex design challenges.

Practically speaking, these findings hold potential for broader application in game development and other fields requiring automatic generation of complex systems or structures. As the mechanics of AI-driven content generation continue to evolve, the integration of such frameworks could enhance both the functionality and aesthetic diversity of machine-generated content.

Looking to the future, the expansion of this research may consider alternative encodings and block sets, the integration of user-interaction within the evolutionary loop, and exploration of cooperative dynamics in multi-agent systems within Minecraft. Further investigation could also focus on other complex outcomes in different digital or robotic domains utilizing similar methodologies.

This work pioneers a previously unexplored niche, showcasing the utility and potential of evolutionary computational algorithms in dynamically challenging and open-ended environments such as Minecraft.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Alejandro Medina (1 paper)
  2. Melanie Richey (1 paper)
  3. Mark Mueller (5 papers)
  4. Jacob Schrum (11 papers)
Citations (2)
Youtube Logo Streamline Icon: https://streamlinehq.com