Site navigation

Assessed project · Class 12

Reinforcement Learning Studio

Watch an agent learn from reward, one action at a time

About 45 minutesAdvancedFrontier Labs

What your child can do at the end

Explain the reward, state, action and Q-value of a reinforcement learning agent

The brief

Nobody told AlphaGo which moves were good. It played, it was rewarded when it won, and it slowly learned which positions lead to reward. In this studio you build the world: walls, a goal that pays, traps that hurt, ice that makes the floor slippery. A tiny agent starts knowing nothing. Every action it writes one number into its Q-table: how good that action was from that square. You control how fast it learns, how far ahead it values reward, how curious it is, and whether it learns with Q-learning or SARSA. You see every value, every arrow and every episode's reward, and a narrator explains why the agent is improving, stuck, or lost.

Skills it develops

reinforcement learningalgorithmic thinking

How it is assessed

3 graded challenges inside the lab, and 8 questions at the end that check the idea rather than the clicks. Every attempt is saved, so a retry after a mistake counts as the learning it is.