Skip to content

Latest commit

 

History

History
16 lines (12 loc) · 1.19 KB

README.md

File metadata and controls

16 lines (12 loc) · 1.19 KB
Raw

Fast_and_Curious_AI_Environment

A collection of Environments to Create, Train and Test different kinds of Artificial Intelligente algorithms to be used inside the Fast and Curious game, which you can find here.

Environment 1 -> Car-Avoiding-Obstacles:

Introduction:

Inside some Missions of the game, there is a Car trying to run away from a Space-Time entity which wants to destroy it. The road is filled with various types of obstacles (which cause harm on impact) and collectables (the more collected the better).

Goal:

Make the Car avoid ALL obstacles while collecting as many collectables as possible.

Approach 1:

  • Model: Neural Networks;
  • Training Algorithm: NEAT;
  • Inputs: x and y position of the car, x and y distance to the 2 closest obstacles, x and y distance to the closest collectable;
  • Results: Goal is easly achieved in under 5 minutes in terms of avoiding obstacles, but it could be improved in terms of collecting items. The algorithm was ran just for a maximum of 20 minutes, which means a better result could be achieved for longer training times.