Landing rockets using Neural Networks and Genetic Algorithms

An exploration on genetic algorithms.
We create a simple neural network from scratch and see if we can make it learn the laws of physics using an evolutionary process.

0:00 intro
0:19 experiment setup
0:40 manual landing
0:52 neural network
1:29 genetic algorithm
1:46 evolution
3:11 different solutions
4:01 different situations

[sublucid]

Very cool! I love it when genetic algorithms train neural networks!

[k.mahesh]

One question, are the inputs and outputs in NN, are they raw numbers, or have they been normalised /scaled in some way to train NN faster?

[igorp]

Really impressive! Could you tell what software are you using to make this beautiful visualization?

[seklerek]

It would be cool to implement limited fuel for this and see how it adapts to it.

[monalisa]

Hello, can we use the real parameter values ​​of the rocket in this simulation? And can we analyze with this system instead of Matlab?

[yourfriend]

Can I know the name of the software that is used here, is there any tutorial available for this :)