Project information

• Engine: Unity
• Team: Mario Galan Sobrino
• Features: 2D Wave Function Collapse algorithm, machine learning, agents
• Project date: April 2023 - May 2023

This project was developed for the AI subject during my final Computing HND year. It was built using Unity and features two different AI techniques.

The techniques is question are Wave Function Collapse or WFC for short and machine learning agents. Regarding the WFC algorithm, it creates procedurally generated racing tracks for the agents to race on. The algorithm receives a bunch of inputs being those all the possible track parts and then, with some rules of which parts and which not can be together, it creates different tracks with every iteration. The size of the track can also be increase given that it uses a grid as the base. Moving into the machine learning agents, I used Unity's machine learning agents toolkit as the base and then modify them for the purpose of my project. For them to learn I added checkpoints to the track that gave them rewards depending on how many they collected. The more rewards they got the better since it meant that the agent was moving along the track correctly.

I enjoyed this project a lot since it allowed me to experiment with tech that I had been very fond to try for some time. The result was great although not as good as I would have hoped for due to lack of time.