Not signed in (Sign In)

Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.

    •  
      CommentAuthorTheDudeFromCI (Advanced Member)
    • CommentTimeMar 1st 2016
     
    As most of you may know, the physics for CIE aren't the best. They tend to be pretty glitchy, under many conditions. Like broken joints, or objects passing through each other.

    Example:


    However, by turning the FPS to 100, and simulation length to 0.01, something interesting happens.


    The effect, though still present, is much much smaller then previous, and more accurate.

    Hope this helps. ^^-----------------
    Orange is my favorite number.
    •  
      CommentAuthorTheDudeFromCI (Advanced Member)
    • CommentTimeMar 1st 2016
     
    Oh! Also, notice how the balls take different paths, even though I changed nothing on the field itself. This leads me to believe that lag is one of the things that create variation between playthroughs on certain designs.-----------------
    Orange is my favorite number.
    •  
      CommentAuthorMathdude314 (Advanced Member)
    • CommentTimeMar 2nd 2016
     
    It is not lag. It is the fact that inaccuracy of the engine creates subtle differences.-----------------
    If less people are active, that will only make for even less activity. Start making those designs!
    Currently working on a one-round no-elimination quick RP game...
    •  
      CommentAuthorTheDudeFromCI (Advanced Member)
    • CommentTimeMar 2nd 2016
     
    Yes, I mean what's causing those subtle differences. If the inputs do not change, the floating point errors will not change.-----------------
    Orange is my favorite number.
    •  
      CommentAuthorMathdude314 (Advanced Member)
    • CommentTimeMar 2nd 2016
     
    Certainly there is a difference between attempting to model precise and chaotic systems at a high seconds-per-step ratio and a low one. Here seconds per step is Phyard seconds not real seconds.-----------------
    If less people are active, that will only make for even less activity. Start making those designs!
    Currently working on a one-round no-elimination quick RP game...
    •  
      CommentAuthorTheDudeFromCI (Advanced Member)
    • CommentTimeMar 3rd 2016
     
    Oh, I know that, obviously. But when I've worked using certain physics engines in game development, sometimes they'll skip steps in the event of lag in order to catch up. Though the steps were often small, and had little noticeable effect to the user, these would eventually add up in larger simulations. This is one of the reasons that physics based cut scenes in games tend be be pre-recorded, and only "look' like real time physics. (The other reason being to save CPU on weaker machines.)-----------------
    Orange is my favorite number.
    • CommentAuthorXyuzhg (Moderator)
    • CommentTimeMar 3rd 2016 edited
     
    I have no idea why you're fixated on lag when there's no sophisticated step skipping going on here. The physics discrepancy is an effect of the change in step size.
    Particularly, the Box2D collision model approximates collisions by applying an impulse in an appropriate direction once two objects begin to overlap.
    Keep in mind that even in the best circumstances this is simply an approximation of what happens in reality, as things aren't perfectly rigid in the real world.

    If your step time resolution is small, then clearly collisions will be detected with better accuracy and you'll get more realistic estimates of the collision impulse.
    (For more detailed information about the computation read here: http://www.iforce2d.net/b2dtut/colli... )

    On a side note Box2D also uses a semi-implicit Euler method which is known to be quite unstable for large step intervals, which might also influence the nature of the simulation.

    Maybe you're getting simulation steps confused with simulation frames?-----------------
    Hopefully PA is inconsistent.
    •  
      CommentAuthorTheDudeFromCI (Advanced Member)
    • CommentTimeMar 3rd 2016
     
    Fine, so how would smaller steps become available? I remember old versions of the editor that had significantly more stable physics.-----------------
    Orange is my favorite number.
    • CommentAuthorXyuzhg (Moderator)
    • CommentTimeMar 3rd 2016
     
    You specify the step size with the "Simulation Step Time Length" simulation setting, which you presumably changed to 0.01 yourself.
    This won't vary after you've set it, unless you use the API that does so.

    For the decline of the physics stability, this is a problem with the update in the Box2D implementation a few years ago.
    Refer to Tapir's last post in this thread: http://forum.colorinfection.com/comm...
    My guess is that they reduced the default number of position/velocity iterations, thereby losing collision detection accuracy.
    Or maybe they changed their integrator. There are quite a few possible reasons for the physics change.-----------------
    Hopefully PA is inconsistent.