A small 2D simulation in which cars learn to maneuver through a course by themselves, using a neural network and evolutionary algorithms.

Interested in how Neural Networks work? Have a look at my one-minute-explanation: automobilrevue.net/v/ckVEelVUM3ltdzQ

This simulation was implemented in Unity. You can find detailed information about how this simulation works, as well as a link to the entire source code on my website: arztsamuel.github.io/en/projects/unity/deepCars/deepCars.html

Don't miss any future videos, by subscribing to my channel.
Follow me on Twitter: twitter.com/SamuelArzt


    • Samuel Arzt
      Samuel Arzt

      In case you are interested in how Neural Networks work, I made a one-minute explainer: https://www.youtube.com/watch?v=rEDzUT3ymw4

      about 1 year ago
    • Alexandr Komagorov
      Alexandr Komagorov In reply to Samuel Arzt

      Could you explain on how to create this kind of environment and how to train dnn on this environment ? Thanks 🙏🏻

      about 1 month ago
    • Thiago Villa
      Thiago Villa In reply to Samuel Arzt

      @Christopher Hogben Wait, what?

      about 2 months ago
    • HR_
      HR_ In reply to Samuel Arzt

      How did you make the simulation?

      about 4 months ago
    • Trung Ng
      Trung Ng In reply to Samuel Arzt

      Is it really DL?

      about 4 months ago
    • Dark Knight
      Dark Knight In reply to Samuel Arzt

      What are inputs ?

      about 4 months ago
    • june

      지랄 말이되냐 왜 계속 나머지는 뒤로 쳐박는건데

      about 2 days ago
    • illigarium

      You should try this with more neurons and with different maps.

      about 2 days ago
    • 귀엽지사막여우

      아 이색기 졸라못하네

      about 3 days ago
    • Bud Yeddi
      Bud Yeddi


      Thx YT recommended

      about 3 days ago
    • Zach Hoffa
      Zach Hoffa

      So this is how Elon developed it!

      about 3 days ago
    • 이재현

      존나 멍청하네 뭐

      about 3 days ago
    • 김동준

      That's how I used to travel my mom's body

      about 4 days ago
    • Desheng Li
      Desheng Li

      A simple vfh algorithm can finish this course. Why brother deep learning

      about 6 days ago
    • Samuel Arzt
      Samuel Arzt In reply to Desheng Li

      Of course, one could even argue that some simple if-else statements can solve this course.

      This project was never meant to serve a true scientific or economic purpose. It was rather meant as a means of education / entertainment. It helped me wrap my head around neural networks back then and the resulting video seemed quite entertaining to watch.

      about 5 days ago
    • Kristaps Smilga
      Kristaps Smilga

      Lol what happened with generation 4. Lol

      about 7 days ago
    • Albert Alfrianta
      Albert Alfrianta

      Impressive! Inspiring!

      about 8 days ago
    • Outside Plays
      Outside Plays

      I feel frustrated every time they repeat the same mistake I mean... It's a damn neural network, it shouldn't repeat mistakes twice!!!

      about 9 days ago
    • Samuel Arzt
      Samuel Arzt In reply to Outside Plays

      You're right, a Neural Network should be (somewhat) deterministic and therefore not repeat mistakes that it has already overcome. The cars that are falling for mistakes of previous generations are however mutated individuals of the current best. Without mutation, there would be no exploration and therefore no improvements. So the Genetic Algorithm is actually to blame here, not the Neural Network.

      about 9 days ago
    • Ahammed Shamir
      Ahammed Shamir

      Could you please help me learn how to do this? I am new to ML but know python. I am baffled by genetic algo but has no clue how to do. Do u have a tutorial or some links that you can help with.

      Amazing video.

      about 9 days ago
    • Samuel Arzt
      Samuel Arzt In reply to Ahammed Shamir

      @Ahammed Shamir Yes, I mainly used Unity's built in components / mechanisms for the simulation and wrote custom C# scripts for the Neural Network and Genetic Algorithm.
      Nowadays there is also the ML-Agents framework developed by unity which goes beyond evolutionary algorithms and sets you up with deep reinforcement learning.

      about 9 days ago
    • Ahammed Shamir
      Ahammed Shamir In reply to Ahammed Shamir

      @Samuel Arzt thanks for such detailed reply. I will surely check that paper. So you mean to say the whole thing is done in Unity?

      about 9 days ago
    • Samuel Arzt
      Samuel Arzt In reply to Ahammed Shamir

      Thanks for the kind words!

      Python is a very good choice of language for machine learning and data analysis in general. Unfortunately I can't really point you to any specific tutorials for a genetic algorithm in python, but I am sure there are many out there. I think it is great to program such algorithms from scratch if you want to truly understand them, however you have to be aware that for real world practical applications of machine learning / data analysis you should be using one of the many existing libraries. Python has a load of them, which is also one of the main reasons why it became so popular in the field.

      I personally have learned about the Genetic Algorithm through a paper by Darrell Whitley ("A Genetic Algorithm Tutorial"). It is quite long but you don't have to read the entire thing to understand it, just two or three of the main sections.

      The entire source of this project, which was made in Unity, is also available on GitHub (see link in video description) maybe that is helpful to you.

      Hope that helps you in some way!

      about 9 days ago
    • ezic

      https://www.youtube.com/watch?v=gnfkfUQvKDw :D

      about 12 days ago
    • Eduardo F.N.S
      Eduardo F.N.S In reply to ezic

      eu pensei exatamente nisso quando vi que esse vídeo é de dois anos atrás

      about 6 days ago
    • 생각의자

      same like human.

      about 12 days ago
    • Si a
      Si a


      about 17 days ago
    • UberSoldat 05
      UberSoldat 05

      Computer deep learning how to drive,
      It is time for learning how to sex.

      about 17 days ago
    • 414 435 43C 44C 44F 43D 20
      414 435 43C 44C 44F 43D 20

      Ничего не понятно, но очень интересно

      about 23 days ago
    • Jon Iwanyszyn
      Jon Iwanyszyn

      We have to break the signal...the signal must be cut off at the source...

      -They Live Documentary

      about 1 month ago
    • Dave B
      Dave B

      Amazingly enjoyable

      about 1 month ago
    • Bobby childs Lee
      Bobby childs Lee

      Oddly interesting.. Idk what it but..🙇

      about 1 month ago
    • Warlock

      Это точно нейронка? Может генетический алгоритм?

      about 1 month ago
    • Nutan Yadav
      Nutan Yadav

      What programming language was this coded in?

      about 1 month ago
    • H R
      H R

      is this how sperm eventually reaches the egg?

      about 1 month ago
    • WayoftheDave

      I can't wait for this technology to be used in real cars, after the initial body count, this will be way better

      about 1 month ago
    • Zappy Skrrt
      Zappy Skrrt

      But for this to be actually learning to drive wouldn’t it have to be on randomizes tracks, or at least multiple, because right not they are learning exact movements for this track and not necessarily learning the rules of driving

      about 1 month ago
    • Lon Link
      Lon Link

      That's how babies are made

      about 1 month ago
    • Mister We
      Mister We

      Shit why i can't understand this beauty 😭

      about 1 month ago
    • Shitpostin Is Awesome
      Shitpostin Is Awesome

      if you add enviromental variables such as rain and snow, would they be able to decide whether or not speed?

      about 1 month ago
    • リョウギョテツ

      could you please tell me what language did you use?

      about 1 month ago
    • リョウギョテツ
      リョウギョテツ In reply to リョウギョテツ

      @Samuel Arzt Thank you so much for replying.

      about 1 month ago
    • Samuel Arzt
      Samuel Arzt In reply to リョウギョテツ

      I used C# with the Unity game engine.

      about 1 month ago