Interactive lab
Watch a tabular reinforcement-learning agent learn a safe path through a maze.
Watch a tabular reinforcement-learning agent learn a safe path through a small 6×6 maze.
Episode
0
Total reward
0.0
Status
Idle
Training is manual. The loop pauses when this panel is off-screen or the tab is hidden.
Last 100 completed episodes.
Q-learning stores a score for every action in every square. The agent explores, receives a reward, and improves its choices over many episodes.
α controls how quickly new experience changes a score. γ is the 0.9 discount for future rewards. ε controls exploration and gradually decays to 0.05. This demo implements tabular Q-learning; PPO and SAC are separate algorithms for continuous-control problems and are not implemented here.
Interactive lab
Watch a tabular reinforcement-learning agent learn a safe path through a maze.
Watch a tabular reinforcement-learning agent learn a safe path through a small 6×6 maze.
Episode
0
Total reward
0.0
Status
Idle
Training is manual. The loop pauses when this panel is off-screen or the tab is hidden.
Last 100 completed episodes.
Q-learning stores a score for every action in every square. The agent explores, receives a reward, and improves its choices over many episodes.
α controls how quickly new experience changes a score. γ is the 0.9 discount for future rewards. ε controls exploration and gradually decays to 0.05. This demo implements tabular Q-learning; PPO and SAC are separate algorithms for continuous-control problems and are not implemented here.
