What machine learning methods can be used to train a control system on a fitness function other than a genetic algorithm?

Question

Neural networks

Task:
The car drives along the track, collecting bonuses along the way. At the input we have data, a map of the area within the field of view, current speed, current position, at the output of the control system commands: when and how to turn, when to accelerate, when to slow down, etc. The fitness function evaluates the optimality of the distance traveled.
What machine methods can be used to solve this problem, except for genetic algorithms for a neural network?

Answer 1

Q-algorithm
And the successful use of neural networks to approximate the table Q(S, a), implemented by Deep Mind in the form of DQN https://arxiv.org/abs/1312.5602
After that, there were many more different articles that solved all sorts of problematic points like actions with parameters. The last thing I saw on this topic is the A3C architecture https://arxiv.org/pdf/1602.01783.pdf
On the Internet, you can find posts with implementations in different languages ​​and libraries.

Answer 2

It looks like a task for reinforcement learning. Read, for example, karpathy.github.io/2016/05/31/rl