Automatic Map Generation

Prediction, planning


  1. Process historical F1 data
    • Look at the racing lines that these drivers take for qualifying, as well as the velocity
  2. Frame it as a supervised learning problem, based on the racing lines that the driver followed the previous day.

But the thing is with racing, you want to maximize speed.

Another thing is to be able to run this real-time. And solving things are very slow.

With machine learning, you can just do inference.

Two parts:

  • Prediction, and then planning

This idea of encoding racecraft into robots. End goal: Go-kart racing against a human driver.

  • Dynamics of the vehicle