REINFORCEMENT LEARNING · RUNS IN YOUR BROWSER

Build the brain
of a self‑driving car.

A one-hour, hands-on reinforcement learning workshop that runs entirely in your browser. Write Python. Train a real Q-learning agent. Watch it measurably improve — nothing to install.

5 Steps 60 Minutes Q-Learning No Install

Get Started View Mission Brief

LIVE DEMO

training.log ● LIVE

Ep 001 Reward −82 Dist 14m

Ep 010 Reward −21 Dist 38m

Ep 050 Reward +132 Dist 124m

Ep 100 Reward +287 Dist 312m

Ep 300 Reward +641 Dist 748m

Ep 500 Reward +1204 Dist Complete

agent improving_

Agent XP 2,847 / 5,000

Four missions.
One trained agent.

Each mission builds on the last. Start with two lines of Python — finish with a self-trained autonomous agent.

01 Entry

Control the Car

Return one action string. Watch the car move. Understand that your code is the agent's entire brain.

Write choose_action()

Return "ACCELERATE"

See the car respond

XP Reward +50 XP

02 Sensor

Read the Sensors

Use distance readings and track angle to make decisions. State → Action in its purest form.

Read front_distance

Write conditional logic

Avoid walls, reach checkpoints

XP Reward +150 XP

03 Reward

Design Rewards

Define what "good" and "bad" mean for your agent. The AI optimises for exactly what you reward.

Edit reward.py

Tune checkpoint bonuses

Watch behaviour change

XP Reward +300 XP

04 Final

Train and Compete

Run hundreds of episodes. Watch reward charts climb. Replay before vs after. Top the leaderboard.

Run 500 episodes

Compare before vs after

Top the leaderboard

XP Reward +500 XP

The loop that drives every AI.

Every autonomous agent — from your Q-learning car to the latest language model — runs this same cycle. You'll experience it live.

Observe State

Speed, distances, track angle — 6 numbers

Choose Action

Your agent decides: ACCELERATE, TURN, BRAKE

Environment Changes

The car moves, physics update

Receive Reward

+1 per meter, +50 checkpoint, −100 crash

Learn and Improve

Q-table updates, policy gets smarter

student/agent.py only file you edit

# ── STUDENT AREA ──────────────────────
# Write your agent logic here

def choose_action(state):

    speed = state["speed"]
    front = state["front_distance"]
    left  = state["left_distance"]
    right = state["right_distance"]

    if front < 30:
        return "BRAKE"
    if left < 15:
        return "TURN_RIGHT"
    if right < 15:
        return "TURN_LEFT"

    return "ACCELERATE"

Train the best agent.
Top the leaderboard.

At the end of the workshop, every team's trained agent is scored on track completion, reward, and driving efficiency. The leaderboard updates live.

Scored on completion, reward, collisions, and time
Local leaderboard — no accounts, no internet required
Replay any team's agent to see what it learned

Leaderboard Live

1 Team Alpha 100% 982

2 Team Beta 91% 914

3 Team Sigma 87% 887

4 Your team... — —

Score: completion 40% · reward 35% · collisions 25%

Build the brainof a self‑driving car.

Four missions.One trained agent.