#17 Some type of noise to sensors/movement

Next_release
closed
5
2007-09-11
2004-01-15
Anonymous
No

I know that this idea has been suggested before, but it
is really necessary for some types of robot controllers
to have something less than perfect sensors/movement.
Even adding a bit of randomness, any kind, would be
better than the perfect world. Of course, it should be
able to be turned off, too. It really doesn't have to
have a model behind it. It could be a type of callback
function, so that we define our own. The simulator that
comes with Saphira/Aria does a good job, if you need
some ideas.

Discussion

  • Richard Vaughan

    Richard Vaughan - 2004-02-13

    Logged In: YES
    user_id=139639

    Noise models are in the pipeline. I'm trying to do this in a modular way,
    so it's easy to create your own models and attach them to devices. Look
    for some basic noise stuff in 1.4.X when it arrives. I'm aiming for mid-
    march.
    RTV

     
  • codohundo

    codohundo - 2004-10-24

    Logged In: YES
    user_id=1145267

    Forgive me if this is the wrong place to post this, I'm new
    to SourceForge. This would be great feature, I was hoping to
    use this software for an important (that is to say important
    to me) project. Is there any news on this feature and is
    there anything I can do to help?
    codohundo

     
  • Reed Hedges

    Reed Hedges - 2005-05-03

    Logged In: YES
    user_id=39867

    Odometry drift is the most important aspect here, I think.

     
  • Reed Hedges

    Reed Hedges - 2005-12-14

    Logged In: YES
    user_id=39867

    Odometry error has been implemented in 2.0.0a.

    I have patches to do basic noise in for the laser and
    ranger, but it's only customizable by changing a constant
    (for max. error variance) in the world file, you can't
    implement your own noise function, which I think would be
    the ultimate goal, right?

     
  • Richard Vaughan

    Richard Vaughan - 2007-09-11
    • status: open --> closed
     
  • Richard Vaughan

    Richard Vaughan - 2007-09-11

    Logged In: YES
    user_id=139639
    Originator: NO

    Odometry error is now implemented.

    My philosophy on sensor noise is that it's not necessary to simulate it. The stage ranger models are already bad, in that they don't closely model any real-world device. That badness means you should not assume that your real robot sensor values will be like Stage's values. You can't take Stage's values seriously, except in statistical sense. Which is exactly where you are with sensor noise. So why take the computational cost of adding noise? I think that most people find this is a reasonable argument, and so for now I am not adding sensor models to Stage. Of course, if you disagree, I welcome your noise-inducing patches.

    - rtv

     

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks