Using NEAT(Neuroevolution of Augmenting Topologies), a simulated agent learns to drive a car through a series of race maps by mutating and adopting the 'genotypes' of 'parents'. The environment is set up in such a way where the agent uses multiple copies of itself to explore the environment. The copies that do not progress further are eliminated. This process repeats through a number of generations, until the agent is able to navigate through this environment as fast as possible
The Algorithm used in the program is NeuroEvolutionary Augmented Topologies [N.E.A.T]
Refer the paper here: http://nn.cs.utexas.edu/downloads/papers/stanley.ec02.pdf
Mutational Diagram:
