Designing a Video Game AI with Reinforcement Learning
summary
Design of a video-game AI for a turn-based flicking/puck game using reinforcement learning, implemented with an Actor-Critic architecture (two neural networks acting as player and coach) to handle continuous actions (force, direction, rotation). The talk covers RL fundamentals for games, why Actor-Critic suits continuous-action games, technical implementation details using TensorFlow.js and Playwright for automation, training challenges, results, game integration, and practical lessons for developers.
AI-written summary · about
related talks
02mDesigning a Video Game AI
31mDesigning game AI with reinforcement learning
22mDeveloping AI in Racing Games Using Reinforcement Learning
56mCan I do AI?
52mThe Journey of AI in Games: A Technical and Philosophical overview
30mReinforcement Learning - ChatGPT, Playing Games & More
01h 01mIntroduction to Practical AI
28mDeep Reinforcement Learning: theory, intuition, code