Autonomous Agents Evaluate Procedural Content Generation in Endless Runner Game
A recent study on arXiv introduces Momentum, an endless-runner game that combines runtime terrain generation, environmental object spawning, and autonomous agent evaluation into one cohesive gameplay experience. As players progress, ground tiles and environmental elements are generated in real-time, with their placement governed by a constraint-driven method inspired by Wave Function Collapse (WFC). The navigation surface is asynchronously rebuilt to align with the streamed environment. Two autonomous agents scout ahead: an aerial scanner analyzes the corridor's geometry, while a ground-traversal agent checks the path for solvability. This method tackles challenges in Procedural Content Generation (PCG), where generated content can be unbalanced, obstructed, repetitive, or unsolvable. The paper can be found on arXiv with the identifier 2605.01783.
Key facts
- Paper titled 'Runtime Evaluation of Procedural Content Generation in an Endless Runner Game Using Autonomous Agents'
- Published on arXiv with identifier 2605.01783
- Introduces Momentum, an endless-runner game
- Uses runtime terrain generation and environment object spawning
- Object placement inspired by Wave Function Collapse (WFC)
- Two autonomous evaluation agents: aerial scanner and ground-traversal agent
- Addresses PCG evaluation problem of unbalanced or unsolvable content
- Navigation surface rebuilt asynchronously
Entities
Institutions
- arXiv