Maze Q-Table

Reinforcement learning allows an agent to learn through trial and error in an environment. Imagine a maze. Using a Q-table, the agent explores the maze (state) and takes different actions (like moving up, down, left, or right). The Q-table keeps track of the rewards (positive for good choices, negative for bad decisions) received for each action in each state. Over time, the agent updates the Q-table to prioritize actions that lead to the highest reward and eventually learns the best path through the maze.

Learn by Doing

In this app, you'll guide an agent through a maze filled with bricks. The agent will use a technique called Q-learning to navigate the maze and reach the goal. By observing the agent's exploration and learning process, you'll gain an understanding of reinforcement learning concepts.

*Adjustable Difficulty

Want to make things easier or harder for the agent? You can control the difficulty of the maze by adjusting two settings:

Brick Obstacle Size: Choose the size of the bricks that the agent needs to navigate around. Larger bricks create a more open maze, while smaller bricks make for a tighter squeeze.

Number of Episodes: This setting determines how many times the agent will attempt to learn the maze. With more episodes, the agent has more opportunities to explore and refine its decision-making strategy.

Take Control and Track Progress

The app provides two buttons to control the simulation:

Generate Maze: This button creates a new maze with the chosen brick size.

Start Learning: Once you're happy with the maze layout, click this button to start the agent's learning process.

Visualize Learning Progress

In addition to observing the agent navigate the maze, you can also track its learning progress with the included Learning Rate Graph. This graph plots the agent's Rewards (positive reinforcement) versus the number of Episodes. By analyzing this graph, you can see how the agent's learning rate impacts its decision-making and how it improves over time.

Get ready to witness the power of reinforcement learning firsthand! As the agent explores the maze, you'll see how it learns to make optimal choices and conquer the challenge. The learning rate graph will further enhance your understanding by visualizing the agent's progress throughout its training journey.