A 2D Unity simulation in which cars learn to navigate themselves through different courses. The cars are steered by a feedforward neural network. The weights of the network are trained using a modified genetic algorithm.
If I train the cars in one track and then use those cars on a new track, will they be able to succeed in first attempt or they have to relearn the new track ?
Hey there id like to get the most recent weights values after closing the simulation and save these to a text file so i can use them on the next time i open the simulation. Could you assist me with this? great repository by the way!
You are most familiar with your code. Can you see a strait forward way to change the map at certain intervals or perhaps after a certain goal is achieved to perhaps train the model to become more adaptive. Perhaps this already occurs, or does the model just learn to succeed at one specific map at a time?