Summary: Training AI Bots in DDNet: A Journey in Reinforcement Learning
Dive into my journey of training 1,500 parallel bots to conquer DDNet, a cooperative 2D platformer. This project is a fusion of passion and technology, rooted in my love for gaming and AI.
Highlights:
- Concept Origin: Inspired by SethBling’s MarI/O and Yosh’s RL agent.
- Tech Tools: Utilized C++ and Python for neural network training with ONNX Runtime for efficient processing.
- Learning Mechanism: Bots are rewarded for their speed and strategy in completing levels, revealing both impressive tactics and amusing failures.
Key Takeaways:
- Iterative Process: Faced challenges in developing a robust reward system and debugging complex neural networks.
- Community Support: Engaged with the DDNet community for insights into mapping intricacies.
- Future Plans: Aim to implement curriculum learning for diverse map training and explore various game modes.
Join me as I push the boundaries of AI in gaming! 💡✨
Let’s connect and discuss AI innovations—share your thoughts below!
