Contents
This is a course for students interested in methods of procedural content generation (PCG) in computer games. An overview of PCG methods is presented and a variety of approaches is investigated more thoroughly. The labs focus on the practical usage of PCG in more-or-less developed computer games and use some elementary knowledge of Java/Kotlin and Python (no prior knowledge of the languages required, with willingness you can learn as you go). The course is taught at MFF UK as NCGD011.
News
Follow the appropriate channel at Gamedev Discord!
https://discord.gg/Zts98PGw6z
Dates (SIS)
Lectures will be interwoven with Labs on even / odd weeks, but the course starts with two consecutive lectures.
Course Exam
There will be an oral examination done during the examination period, either in-person or online, depending on the situation. The exam will be a mixture of direct question on the topics below and open questions where you will be describing a procedural approach for given content.
Exam topics
Classification of PCG: Reasons for PCG in games. Design-time vs runtime. Teleological vs ontogenetic. Direct vs indirect.
Terrain: Simple approaches for terrain generation. The diamond-square algorithm.
Noise: Value vs. gradient noise. Octaves. Perlin noise. Advantages of simplex noise. Domain warping. Use-cases for analytical derivatives. Billow and ridge noise. Cellular (Worley) noise. 3D approaches. Perlin worms.
Search-based: Pros/cons of various content representations. Building blocks. Template search. Evolutionary algorithms – differences when using for PCG. Types of evaluation functions. Assessing quality of generators.
Methods: Cellular automata. L-systems. Shape and graph grammars. (watch out, is in art and music bonus lecture)
Visual art: Art toys (what are they, what are they for). Procedural effects.
Musical art: Approaches for music. Melody. Harmonization.
Constraint-based programming: Answer set programming – description, use-cases. AnsProlog. Wave function collapse.
Mixed-initiative approaches: Definition. Three initiatives (of dialogue). Pros/cons.
Mazes: Types. Attributes. Objectives. Solving (computer/human methods). Perfect maze generation (Kruskal’s algorithm, backtracking).
Dungeons: Combining maze algorithms with rooms. Space division methods. Mission graphs + space layout options. Enemy placement. Spawning waves.
Loot: Methods. Affixes. Loot tables. Variability. Player interaction with loot system.
Puzzles: Approaches.
Machine Learning: Types of learning, GANs, latent space. Outline of LLMs, advantages, disadvantages for use in PCG.
Lectures
Lectures Schedule
| No. | Date | Topic | Lecturer | Content | Slides |
| 0 | 20.2. | Introduction | Vojtěch Černý | What is Procedural Content Generation Reason to use PCG Brief history Interesting implementations Course structure |
PPTX, PDF |
| 1 | 27.2. | Terrain | Vojtěch Černý | Why generate terrain Simple approaches Noise functions 2D and 3D approaches |
PPTX, PDF |
| 2 | 13.3. | Search-based Approach | Vojtěch Černý | Search-based approach Content representation Search algorithm Evaluation & Quality Case Study: Spelunky & Yavalath |
PPTX, PDF |
| bonus | – | Art and Music | Vojtěch Černý | self-study material Art toys How to do procedural effects Layman’s approach to music generation |
Lecture PPTX, Lecture PDF, (voluntary) Labs PPTX, Labs PDF |
| 3 | 27.3. | Constrained PCG | Vojtěch Černý | Constraint-based approach Answer Set Programming Wave Function Collapse |
PPTX, PDF |
| 4 | 17.4. | Mazes and Dungeons | Vojtěch Černý | Generating Mazes and Dungeons Procedural Loot Wave Spawning Procedural Puzzles |
PPTX, PDF |
| 5 | 15.5. | Deep Learning | Vojtěch Černý | Intro to Machine+Deep Learning and LLMs Advantages / Disadvantages of LLMs for PCG Examples |
PPTX, PDF |
Labs
The labs will consist of short introductions to specific PCG contexts (usually games) and homeworks.
A total of 5 homeworks worth 18 points total will be presented during the semester. 12 points are required for admittance to the exam, and any above 12 will be transferred to the exam points (Exam has a 50 point maximum and 40 (42 from 2027) points are needed to pass, so extra points are significant).
For the 1st practical please have your laptop with preinstalled Java (JDK) 25+. Modify PATH and JAVA_HOME environment variables such that it is your default. Also (for your comfort), IntelliJ IDEA is recommended – community edition is sufficient, but as a student you can get ultimate edition for free – basically by just asking for it on their site.
For the 2nd practical, you will need the setup from above, and also Python 3 installed and added to your PATH.
The 4th practical will be one large homework, worth 6 points. Make sure to allocate more time for it than for others.
Submission details of each HW is outlined in its slides.
Labs Schedule
| No. | Date | Topic | Lecturer | Slides | Prerequisites | Homework | HW pts | Standard Deadline |
| 1 | 4.3. | Minecraft | Vojtěch Černý | PPTX | Java 25, (recommended) IntelliJ IDEA |
HW1 – Minecraft | 3 | 20.3. 23:59 |
| 2 | 20.3. | Mario | Vojtěch Černý | PPTX | Java (25), (recommended) IntelliJ IDEA |
HW2 – Mario | 3 | 10.4. 23:59 |
| 3 | 10.4. | Murder Mystery | Vojtěch Černý | PPTX | Python 3 | HW3 – Murder Mystery | 3 | 24.4. 23:59 |
| 4 | 24.4. | Roguelike | Vojtěch Černý | PPTX | Java 25, (recommended) IntelliJ IDEA |
HW4 – Roguelike | 6 | 15.5. 23:59 |
| 5 | 22.5. | LLMize | Vojtěch Černý | PPTX | some of the above… | HW5 – LLMize | 3 | 10.6. 23:59 |