Providing Rapid Design Feedback for 3D Obstacle Course Games Using Constrained Solvability Queries

arXiv cs.GR details a system for 3D obstacle-course games that gives designers rapid feedback on how a layout can be solved. The tool answers constrained queries such as avoiding regions, passing through waypoints, or limiting jumps, so teams can probe intended routes and edge cases without waiting on slow manual playtests.

The core of the system is a high-performance GoExplore-style search pipeline guided by an offline-trained RL agent, plus a custom GPU-accelerated simulator that can generate playthrough experience at nearly 14,000x real time while still targeting 60-fps gameplay behavior. That combination is what makes the feedback loop practical for iterative level design rather than just offline analysis.

For developers, the interesting part is not only whether an obstacle is solvable, but how it is solvable. The authors say the tool can surface undesirable solution paths, estimate difficulty, and help designers discover new directions they did not originally anticipate. Human playtesting of obstacles built with the system reportedly confirmed that players followed the intended behavior.

The paper also ships code for the interactive tool, simulator, training setup, and a procedural level generation system. That makes it especially relevant for teams building traversal-heavy games, puzzle-platformers, or any workflow where authored challenge needs to be checked against emergent player movement.