yasirrhaq / Learning-to-kart

I tried to train AI to learn how to kart with Reinforcement Learning using UnityML-Agents

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Learning-to-kart

Hi there, I tried to train AI with Reinforcement Learning (RL) using UnityML-Agents. The task AI needed to be done is steering left/right, acceleration, and avoiding obstacles.

Don't know what is UnityML-Agents? Go ahead learn more about it here

mlagent

I also use Karting Microgame from Unity Asset Store that you can get it from here

karting

Main Goal :

  1. AI can learn how to steering left/right
  2. AI can learn how to accelerate
  3. AI can learn how to avoid obstacles
  4. AI can drive in many different tracks without crashing

Rules :

  1. AI will be trained only on one track to learn everything that needed to be done such as steering left/right, accelerate, and avoid obstacles
  2. AI will be tested on different environment such as different track (with and without obstacles)
  3. If AI hit obstacle/track AI position will be reset to the starting point and will be given penalty -1
  4. If AI success reaching checkpoint will be given reward +1

Learning Track

AI WILL ONLY LEARN ON THIS TRACK

trek 3 obstacle

Testing Track

These Are All The Track

all circuit

Training Progress

RL_Training

Result

RL_Demo

About

I tried to train AI to learn how to kart with Reinforcement Learning using UnityML-Agents


Languages

Language:C# 79.8%Language:Mathematica 11.8%Language:ShaderLab 7.4%Language:HLSL 1.1%