Self-Driving car with Neural Network Library 2.0

Help and Feedback Scripting Support
#1

Hey there!
I would like to be able to create cars that drive around the map for my upcoming game that interact with players around them (e.g. if a player car drives into them or a light is red they will stop or avoid a collision

Normally I would use TweenService or PathfindingService but those do not interact with the players around them and cannot tell whether lights are red. I looked up some tutorials but only found videos that used .Touched events on parts to steer the car which I think is highly inefficient. I also am new to the AI scene and have never really toyed with this before

I have tried reading up about Neural Network Library 2.0 on their website but I couldn’t figure out what I needed to change to make something more close to a self driving car.

If you are unsure what the Neural Network Library is, here is a link, https://devforum.roblox.com/t/neural-network-library-2-0/869557

If someone could point me in the right direction with which parameters I should be changing or clue me in on how this works that would be very useful,
Thanks!

#2

what you want to do is have a GA based integral approach where the chromosomes represent each hidden layer in your neural network. Suppose we want two hidden layers for our NN we can represent the chromosome as a integer based representation

chromosome = {0,0}

where chromosome[1] represents the amount of neurons in the first hidden layer and the other the second

To add on you want to loop your randomly generated GAs and forward propogate x generations then see which setup gave u the best results. from there its just inputs of ray distances and position and an output of i suppose a turn speed?