Create your own simple reward game to practice RL ideas.

Here’s a fun project without any coding:

1. Draw a grid map. On a piece of paper, draw a small grid (for example 5×5 squares). Mark a start square and a treasure square (the goal).

2. Place rewards and penalties. In some squares, write a positive point value (like +10) to represent finding treasure or bonus. In other squares, write a negative value (like –5) to represent a trap or penalty. Leave other squares as 0 (no change).

3. Choose a path. Pretend you are an agent moving from Start to Treasure. Pick a route (up, down, left, right moves). For each square you move through, write down its point value. Add them up when you reach the treasure.

4. Try a new route. Now imagine you made different choices (explored a different path). Go from Start to Treasure again via a new path. Write and total the points for this path.

5. Compare your scores. Which path gave you more points? Which actions (moves) led to higher rewards? Discuss why one path was better.

Through this game, you simulate an agent learning the best way to reach the treasure for the most points.

Each play is like one trial where the agent learns how rewarding the path was. Over a few tries, you’ll see which sequence of moves leads to the highest total.

This simple experiment shows the core of RL: trying actions, getting rewards, and finding the best strategy. Enjoy exploring and learning!