There is more to PCG than Meets the Eye: NPC AI, Dynamic Camera, PVS and Lightmaps

TL;DR

This paper explores how advanced procedural content generation techniques in a game develop additional metadata like navigation graphs, voxel spaces, portals, and lightmaps, enhancing AI, camera, PVS, and lighting systems.

Contribution

It demonstrates the integration of diverse PCG-generated metadata into a game engine, improving multiple subsystems with novel procedural approaches.

Findings

Generated navigation graphs for NPC pathfinding across various creature types

Created voxel spaces enabling discrete A* in 3D space for dynamic camera control

Developed procedural lightmaps with ambient occlusion and tessellation for high-quality rendering

Abstract

Procedural content generation (PCG) concerns all sorts of algorithms and tools which automatically produce game content, without requiring manual authoring by game artists. Besides generating com-plex static meshes, the PCG core usually encompasses geometrical information about the game world that can be useful in supporting other critical subsystems of the game engine. We discuss our experi-ence from the development of the iOS game title named "Fallen God: Escape Underworld", and show how our PCG produced extra metadata regarding the game world, in particular: (i) an annotated dun-geon graph to support path finding for NPC AI to attack or avoid the player (working for bipeds, birds, insects and serpents); (ii) a quantized voxel space to allow discrete A* for the dynamic camera system to work in the continuous 3d space; (iii) dungeon portals to support a dynamic PVS; and (iv) procedural ambient occlusion and tessellation of a separate set of simplified meshes to support very-fast and high-quality light mapping.