I have an RC car with a camera, I have implemented so that i can detect lanes on my track (think like a nascar track). I want to get this car to be able to go around the track autonomous. But I am quite unsure what my next step should be.

Either I can do an algorithm that detects so that I stay in the middle of the lane (if i get to close to either of the lines I steer towards the center).

Or perhaps go around the track manually and save the coordinates of the detected lines as well as the actions I take (steering) and try a DQN approach.

I'm trying to minimize my 'trial and error' time a little here. Perhaps there are some important steps in between here, or a solution that I have not thought of (i.e. am I missing something)?

I'm doing this as a proof of concept therefore I can only spend maximum of a month on this, so what would you do here?

  • $\begingroup$ The point of doing something like RL is not necessarily to make the car steer itself - but to demonstrate that a system you have implemented can learn to steer by itself with minimal supervision, or maybe achieve some difficult to pre-program goal such as travel at maximum possible speed whilst still retaining control. A more simple feedback control loop might suffice if your only goal is have the car follow a track that it can reliably detect. Could you give more detail about your goals in the question? $\endgroup$ Mar 29 '19 at 0:02
  • $\begingroup$ @NielSlater Yes, I think that RL might not be the best approach. My goals are quite "simple", i want the car to be able to stay in its lane and be able to make turns. So that it can complete the track. $\endgroup$ Mar 29 '19 at 7:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.