Question on AI learning

Discussion in 'General Discussion' started by basil_fawlty, Jan 29, 2023.

  1. basil_fawlty

    basil_fawlty Registered

    Joined:
    Oct 5, 2021
    Messages:
    31
    Likes Received:
    11
    One thing about rFactor2 which keeps me away from playing is the amount of effort it requires to set up a race vs AI. While in other games you can jump in easily, in this game loading times are long. Ideal difficulty settings vary in a wide range. And we are also supposed to run a whole practice to train the AI for acceptable performance.

    Two things which could make it somewhat easier:
    • What does a 25-30 min practice do in comparison to a longer? My CPU is old and even with accelerated speed it takes some time a 60 min session. Would a shorter one suffice?
    • If I exit and start the event again (for example, I have spent 40 mins practicing only to find out that I needed to adjust diff by 2%), does the learning start from zero again? Same track, same class... It's just incomprehensible why the game cannot store what it has learned.
     
  2. davehenrie

    davehenrie Registered

    Joined:
    Jul 6, 2016
    Messages:
    7,453
    Likes Received:
    4,369
    One of the keys for a better offline experience is ensuring all the AI get completed laps. Now the good new is, with this new release candidate, one of the bugs that prevented all the AI cars from attempting practice & qualifying laps has been fixed. Time isn't important, just making sure the AI do get completed laps.
    But I will offer a partial solution to your problem.
    When you watch a replay, you can jump in at anytime and continue the race from that point. So try this. Run Practice & qualifying, bite the bullet and take the time to make sure all the cars set qualifying times. Then start the race and exit, I'm not sure if editing the AI% will take affect when you reload the replay, but it might be worth testing. If you have replays enabled, this is like a saved game. You can reload the replay at any time pick up the race right there at the start, or half way, whenever you choose.
    ONE of the drawbacks of this method is that you only can use the same cars & track combination for each saved replay. That can get boring. Another drawback is that several times over the past years, improvements in the core code of the game have resulted in older replays no longer working.
     
  3. Manfredk2

    Manfredk2 Guest

    Afair from the past the AI learning (both methods) should store some data under userdata/player/settings or so. Not shure if under the track name or car name.
     
  4. Comante

    Comante Registered

    Joined:
    Nov 20, 2013
    Messages:
    1,694
    Likes Received:
    1,218
    But Ai learning should be initializated somehow or it is supposed to be always ON ?
     
  5. seamount

    seamount Registered

    Joined:
    Feb 8, 2016
    Messages:
    57
    Likes Received:
    59
    may you tell me threads/links for details about "AI learning" or improving AI on a track (at least not having them crashing to others AI or wall)
     

Share This Page