RL Academy 🚀

The Ultimate 3D Reinforcement Learning Playground

Generation 1

AI High Score 0.0s

Survivors 12/12

Human Record 0.0s

Interactive Guide

Reinforcement Learning 101

Slide 1 / 5

Pop Quiz Challenge!

1. Reward Engineering Tool

Change how the AI learns by changing the criteria for its "Treats" (Rewards):

Target Balancing (Default) Rewards are highest when the ball sits exactly at the platform's center point. Edge Rider (Tricky Challenge) Rewards are highest when the ball balances *near the outer edge* of the platform! Roller Coaster (High Velocity) Rewards the ball for moving fast, forcing the AI to develop spiral movements!

2. Evolutionary Parameters

Speed up Simulation Time 1x (Normal)

Synaptic Mutation Rate 15%

Wind Force (Continuous) Medium

Interactive Human Play Mode

Think you can balance better than a neural network? Take the steering wheel of the active inspected platform!

Guided Lab Assignments:

🟢

Lab 1: Perfect Symmetry

Let your AI train to balance a ball for at least 12.0 seconds using the 'Target Balancing' reward model.

In Progress

🔒

Lab 2: Reward Shift

Switch the Reward Designer criteria to Edge Rider and achieve an evolutionary survival score of 8 seconds.

Locked

🔒

Lab 3: Human vs Machine

Activate manual control mode and log a human balancing score of at least 10.0 seconds!

Locked

Pro-Tip: To apply direct sensory wind forces to the ball, tap on its wooden platform directly in the 3D grid sky above!

Neural Inspector

Agent (0)

Learning Status Thinking Live

Current Survival 0.0s

SLIDE OFFSET (DISTANCE TO CENTER) 0.00m

Active pathway signals:

Left: Inputs (Observations) Right: Outputs (Tilt Action Forces)

Line thickness represents connection magnitude. Blue pathways excite actions; Red pathways suppress/inhibit actions!

Mission Accomplished!