Home▸AI & Machine Learning▸Artificial Intelligence Reinforcement Learning Simulation

🧪 Artificial Intelligence Reinforcement Learning Simulation

Watch a reinforcement-learning agent learn to hunt in real time: a predator on an 8x8 grid updates a Q-table with the Bellman equation every step, its evasive prey adapts, and a live value heatmap plus learning curve show the policy converging episode by episode.

AI & Machine Learning2DModerate60 FPS📱 Mobile-adapted

This simulation explores reinforcement learning by putting a tabular Q-learning agent to work on a concrete task: hunting a moving target on an 8×8 grid. The predator has no built-in strategy — it starts with an empty Q-table and discovers a pursuit policy purely from trial, error, and the Bellman update, while an evasive (non-learning) prey tries to stay away. A live heatmap projects the agent's current value estimates onto the grid floor, a learning-curve sparkline tracks steps-to-catch across episodes, and adjustable learning rate, discount factor, exploration rate and an ε-decay toggle let you watch the exploration/exploitation trade-off shape how fast — and how well — the policy converges.

⚙ Under the hood

This simulation explores reinforcement learning algorithms through an interactive environment. Users can observe and potentially influence the learning process of an AI agent as it navigates a simulated task.

Reinforcement LearningAISimulation

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install