A reinforcement learning experiment built using Unity ML-Agents where an AI agent learns behavior through reward-based training inside a custom Unity environment.
- Reinforcement learning environment
- AI agent training using rewards
- Custom Unity scene setup
- ML-Agents integration
- Agent movement and decision making
- Unity Engine
- C#
- Unity ML-Agents
- Python
The AI agent receives rewards and penalties based on its actions. Over multiple training iterations, the agent gradually improves its behavior to achieve better results.
- Reinforcement learning basics
- Unity ML-Agents workflow
- Agent behavior design
- Reward shaping
- AI simulation environments
- More advanced environments
- Multiple AI agents
- Better reward balancing
- Improved training performance
The ML-Agents environment was successfully trained using reinforcement learning.
However, due to a compatibility issue between PyTorch, ML-Agents version, and pip dependencies at the time, the final ONNX export file was not generated correctly.
Despite this, the training process itself ran successfully, and the agent learned and improved behavior during training episodes.
Computer Science Student — Unity & AI Portfolio Project