Exploring Reinforcement Learning- Can AI Learn to Play QWOP? - Digi-Key Electronics

Dive into the world of reinforcement learning through a 26-minute video that explores how AI can learn to play the notoriously difficult game QWOP. Witness friends attempting to master the game's challenging controls before delving into the process of constructing an AI agent using reinforcement learning techniques. Learn about creating custom environments with gymnasium, implementing Proximal Policy Optimization (PPO) algorithms, and utilizing tools like OpenCV and Tesseract for image processing and optical character recognition. Follow along as the AI agent progresses from trial and error to successfully scooting along the ground, reaching the 50-meter mark. Gain insights into potential improvements and limitations of the current approach, and access accompanying written tutorials and code repositories for hands-on learning.

- Intro to QWOP
- QWOP attempts
- Intro to reinforcement learning
- Creating a custom gymnasium environment
- Creating a custom Weights and Biases logger
- Train reinforcement learning agent
- Check the agent’s performance
- Test the agent
- Going further to train a better agent
- Conclusion