Expanding Horizons of Level Diversity via Multi-objective Evolutionary Learning

2025-09-29

AI Summary

Existing game level generation methods evaluate diversity along only a single dimension, failing to simultaneously satisfy requirements concerning content structure, player experience, and playability. Method: We propose a multi-objective evolutionary learning framework that formulates level generation as a joint optimization problem over multiple diversity metrics—including topological dissimilarity, behavioral diversity, and difficulty distribution—and concurrently explores the Pareto front of playability and multi-dimensional diversity during training. By tightly integrating generative models with multi-objective evolutionary algorithms, our approach enables end-to-end co-optimization of diverse dimension-specific metrics. Results: Experiments on the Super Mario benchmark demonstrate significant improvements in structural, behavioral, and player-perceived diversity. Moreover, the framework supports demand-driven customization of generation strategies, establishing a novel paradigm for controllable and interpretable diversity-aware level generation.

Technology Category

• Humans and AI: Game Design — Procedural Content Generation & Storytelling
• Machine Learning: Evolutionary Learning
• Natural Language Processing: Generation

Application Category

• Web Mining and Content Analysis: Web data generation and simulation
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

In recent years, the generation of diverse game levels has gained increasing interest, contributing to a richer and more engaging gaming experience. A number of level diversity metrics have been proposed in literature, which are naturally multi-dimensional, leading to conflicted, complementary, or both relationships among these dimensions. However, existing level generation approaches often fail to comprehensively assess diversity across those dimensions. This paper aims to expand horizons of level diversity by considering multi-dimensional diversity when training generative models. We formulate the model training as a multi-objective learning problem, where each diversity metric is treated as a distinct objective. Furthermore, a multi-objective evolutionary learning framework that optimises multiple diversity metrics simultaneously throughout the model training process is proposed. Our case study on the commonly used benchmark Super Mario Bros. demonstrates that our proposed framework can enhance multi-dimensional diversity and identify a Pareto front of generative models, which provides a range of tradeoffs among playability and two representative diversity metrics, including a content-based one and a player-centered one. Such capability enables decision-makers to make informed choices when selecting generators accommodating a variety of scenarios and the diverse needs of players and designers.

Problem

• Comprehensive assessment of multi-dimensional game level diversity metrics
• Optimizing conflicted diversity objectives during generative model training
• Balancing playability with content-based and player-centered diversity metrics

Innovation

• Multi-objective evolutionary learning optimizes diversity metrics
• Treats each diversity metric as distinct objective
• Identifies Pareto front for playability-diversity tradeoffs