Hello, I am a computer science major hoping to leverage on BeamNG as a sandboxed environment to do AI research. I intend to apply Q-learning to make a race car driver. Q-learning needs the state and a means of inputting controls. State tentatively consist of accelerator, brake, steering, velocity, acceleration, gear, RPM, location, facing. I do not foresee much difficulty in obtaining a stream of these data. However, I do not know how to obtain an egocentric view of the track (the vision for the AI). I hope I can be nudged in the right direction. Thank you!
Please contact us directly, so we can start a discussion via email: https://beamng.gmbh/contact/ - thanks!
