https://github.com/andre-seiji/unity-drifting-ai
AI model trained for driving in Unity ml-agents
https://github.com/andre-seiji/unity-drifting-ai
drift ml-agents unity unity3d
Last synced: 5 months ago
JSON representation
AI model trained for driving in Unity ml-agents
- Host: GitHub
- URL: https://github.com/andre-seiji/unity-drifting-ai
- Owner: Andre-Seiji
- License: mit
- Created: 2024-08-21T20:16:40.000Z (almost 2 years ago)
- Default Branch: main
- Last Pushed: 2024-08-21T21:12:20.000Z (almost 2 years ago)
- Last Synced: 2025-02-13T20:34:03.419Z (over 1 year ago)
- Topics: drift, ml-agents, unity, unity3d
- Language: C#
- Homepage:
- Size: 7.94 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Unity-Drifting-AI

This is an experiment on AI learning to drive using the Unity ML-Agents.
Without giving any instructions, could the AI learn to complete a full lap in 30 seconds while managing a drift handling car?
##
Rules:
- A series of checkpoints (blue) were placed on the track and the agent gets rewarded everytime that goes through the correct checkpoint.
- If the agent goes through the incorrect checkpoint or hit the walls or barriers, it gets punished.
- Each episode lasts 30 seconds.

The difficulty is avoiding the walls while not losing control of the car.

Finally, the AI is able to complete a full lap.
##
Car model: https://sketchfab.com/3d-models/toyota-corolla-ae86-trueno-fe02fba6302e450ea8424591493341ea
Track link: https://assetstore.unity.com/packages/3d/environments/roadways/modular-lowpoly-track-roads-free-205188