Constraint-Based Four-Dimensional Spacetime Blending of Learned Game Mechanics Along Orthogonal Planes
Year: 2026
Type: conference-paper
Abstract: Blending mechanics from different games may provide new gameplay opportunities. In this work, we blend tile-based game mechanics to generate levels for blended games. Mechanics are learned from example playthroughs of each game’s levels as changes in tile configurations. In particular, we show that mechanics learned from 2D games can be blended along different orthogonal planes to create a new 3D game; that is, the mechanics from one game apply along one plane and the mechanics from another game apply along a different plane. We ensure completability of the generated levels by also using temporal constraints to generate a solution along with the level, resulting in a level with its solution being generated as conceptual 4D spacetime block, which we encode in a 2D “expanded filmstrip” representation. We demonstrate the application to several tile-based puzzle games.
Authors: Akshar Vandara, Seth Cooper
Institution: Northeastern University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7171735713