Racer Maker

You are a racing teacher. You're teaching an AI how to drive and win the championship.

Install

This project assumes you have the following installed:

• Ruby 2.4 (with bundler)
• Python 3.6 (with virtualenv)

To get the ruby game setup:

cd game
bundle install
cd ..

To get the python machine learning project setup:

cd machine-learning
pip install -r requirements.txt
cd ..

Usage

cd game
bundle exec ruby game.rb record
# Run a few laps.
# Press Q to quit.

cp *.csv ../machine-learning
cd ../machine-learning
python train.py
mv parameters.json ../game/parameters.json
cd ../game

bundle exec ruby game.rb ai

TODO

Part 1: racing game

• Car
  • Forward motion
  • Turning
  • Braking
  • Reverse
    • Go reverse
    • Steer reverse
• Race track
  • Sprites
    • Road
    • Background
  • Physics via Chipmunk
    • Barriers
    • Tarmac Vs Sandtraps friction
• Gameplay
  • Recognise laps
    • Implement checkpoints
    • Define flag checkpoint
    • Incrementally activate next checkpoint
    • Detect going backwards (and disable checkpoints)
  • Recognise victory (3 laps?)
  • Keep time
• UI
  • Menus
    • Track select
    • Pause menu
    • Game over
  • HUD
    • Speed
    • Lap time
• Graphics
  • Camera
  • Drift marks
  • Dust particles
• Sounds
  • Vroom-vroom
  • Drift screech
  • Brake screech
  • Collision thump

Part 2: AI teaching game

• Sensors
• Connect with tensorflow
  • Via manual process
  • Via automated process
  • Via sockets
• AI learning
  • Mimic
    • Play, train, run
    • Learn on every lap
    • Online learning (learn while running)
  • Pain/happiness learning

Architectural notes

[keyboard] -> [Input] -> [Gameplay] -> [Presenter] -> [window]

When recording:

[keyboard] -> [HumanInput] -> [InputRecorder] -> [Gameplay] ...

When replaying:

[InputReplayer] -> [Gameplay] ...

AI sensors:

... [Gameplay] -> [HumanPresenter] -> [window]
               -> [SensorsPresenter] -> [filesystem]