I just stumbled on this video from 2016 and I thought I would share. It's a small 2D simulation in which cars learn to maneuver through a course by themselves, using a neural network and evolutionary algorithms.
It is probably worth noting that this is a very simple 2D sim designed to illustrate some basic deep learning. Real world driving is obviously a lot more complex. Also, the system gets better by learning from it's mistakes which in this case involve the "cars" crashing a lot in the beginning. But it is a neat little video that illustrates deep learning in a simple case.
Here is a sim from 2019 of an AI learning to park:
Note it takes the AI over 50,000 failed attempts before it's first successful parking. By the 310,000's attempt, it is able to park reliably.
It is probably worth noting that this is a very simple 2D sim designed to illustrate some basic deep learning. Real world driving is obviously a lot more complex. Also, the system gets better by learning from it's mistakes which in this case involve the "cars" crashing a lot in the beginning. But it is a neat little video that illustrates deep learning in a simple case.
Here is a sim from 2019 of an AI learning to park:
Note it takes the AI over 50,000 failed attempts before it's first successful parking. By the 310,000's attempt, it is able to park reliably.
Last edited: