Hello and Welcome to Zubatomic.


We wanted to let you know that some of the projects listed on this website use cookies. These cookies are not edible, and they do basic things such as remembering your log-in data or keep game scores and player rankings. And, very important, our cookies do not share data with any third party, do not collect your personal data, do not track your web surfing, listen to you talking at home, or any other personal stuff so many sites and gadgets today do.

We do not offer the option to personalize the use of cookies as other websites do simply because we use them for a single purpose, which is enabling core functionalities in our projects. That is why, if you accept the use of cookies, you will not put on weight, and everything will run smoothly. So come on, say yes to our cookies!

Car Genetic Algorithm Intelligence
Artificial Inteligence > Car Genetic Algorithm Intelligence
Description
A genetic algorithm (artificial intelligence) that learns, through various genetic evolution generations, to move towards a point, avoiding obstacles.

The idea is that every car has a DNA, which dictates how to will move and the steps it will take. Their fitness function, aka how close they got to the goal, will decide whether they reproduce or die. If they reproduce, the DNA of the child will be slightly altered. According to darwinist theory, "survival of the fittest" should mean that eventually, only cars that reach the goal are left.

Ofcourse, there are very large limitations. For example, if the obstacles blocking the goal are too hard to "guess" the way around after a reasonable number of iterations, the cars will never reach the obstacle. To fix this, the cars would need an actual brain that learns how to "get around obstacles" in general, instead of how to "get around those specific obstacles that the cars were trained under". Nevertheless, this was my first attempt at a self-learning computer so I will make another better one soon.

Developed by Sebi, Jorge
Instructions
Go to the link. Insert "0", for gradual artificial selection, and insert "1" for "half-half" artificial selection. The worst half will be destroyed with this mode. With the other mode, the worst ones have a lower chance of surviving, but they still can, and some better cars will die as well. From experience, we find "1" to be more effective.

Click and drag to create obstacles. Right click to change the target's position.

Some images of the project below

Press on the red arrows to cycle through


-
© 2019, Zubatomic, Inc. All rights reserved. Copyright Sebastiaan Szafir.

Submarine photos credit: U.S. National Archives - Public Domain Archive website