Relational Content Generation
summary
Relational programming uses Answer Set Programming to define and navigate procedural content spaces, implementing choice rules to add possibilities and integrity constraints to carve them away; programs produce multiple stable models (e.g., four outcomes from two optional facts, three when mutual exclusion is added). Implementations use clingo and catSat and prototypes include the Gemini game generator and the Germinate structure editor; results show reliable, controllable generation and bidirectional analysis but suffer grounding/scale limits that make real‑time use difficult.
AI-written summary · about
chapters (21)
- 0:00Start
- 3:28Chris Martens
- 7:21Procedural Content Generation
- 8:54Roguelikes
- 13:47Augmenting Human Creativity
- 16:53Reliability
- 17:18Controllability
- 18:27Creativity and Believability
- 18:51Directed Randomness
- 23:36Evolutionary Algorithms
- 28:13Answer Set Programming
- 29:04Biome Generation
- 31:30Choice Rule
- 33:37Headless Rule
- 36:30Integrity Constraints
- 43:13Gemini
- 44:02Narrative Generation Engine
- 48:31Usability
- 50:41Limiting Actions To Encourage Exploration
- 52:27Structure Editor
- 1:03:40The no Blank Canvas Principle
related talks
16mMulti-stage Relational Programming
47mLogic Programming à la Carte
18mMobius: Synthesizing Relational Queries with Recursive and Invented Predicates
31mDesign Datalog Templates for Synthesizing Bidirectional Programs from Tabular E...
20mSemantic Reification: A New Paradigm for Random Program Generation
19mSome Novel miniKanren Synthesis Tasks
14mRelational Synthesis for Pattern Matching
22mRelational Solver for Java Generics Type System