🧪 Q-Learning Agent Pathfinding Simulation – Exploration Rate

This simulation allows you to experiment with different reinforcement learning algorithms and observe their impact on an agent's ability to learn a complex task within a 2D environment. By adjusting parameters like reward functions and exploration strategies, you can see how these choices affect the agent’s performance and convergence. An epsilon control enables a real epsilon-greedy strategy: random exploratory steps instead of moving straight to the goal. Adds an epsilon control enabling a true epsilon-greedy strategy: random exploratory steps instead of always heading straight to the goal.

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