Bind variables in BFL filter

Hi,

We're trying to use a Bayesian Filter to do some sensor fusion for
estimating the joints positions on a joint based robot (estimated with
different sensors).

We now have two versions of the same filter (one using
ExtendedKalmanFilter, the other one a particle filter: BootstrapFilter)
that converges but haven't yet found a way of binding the BFL state
representation variables. In our case, we'd like to make sure that each
variable of the state stays within the joint limits.

Feel free to ask for more details as I'm not exactly sure what is most
relevant.

Cheers,

Ugo

PS: sorry for cross-posting on ROS-answers
http://answers.ros.org/question/189700/bfl-bind-variables/

Bind variables in BFL filter

Dear Enrico,

Thanks for coming back to me. Yes I meant constraining. Thanks for
confirming what I was slowly getting at!

The best course for me seems to go for a good and precise enough prior +
better pdf for our measurements + less system dispersion, so that our
angles converge in the proper region. Our measurements are quite accurate
and are in the region we expect them to be, so filtering the measurements
themselves should not be needed.

Cheers,

Ugo

On 12 August 2014 12:14, Enrico Di Lello <enrico [dot] dilello [..] ...>
wrote:

> On Tue, Aug 12, 2014 at 12:00 PM, <bfl-request [..] ...>
> wrote:
>
> > Send BFL mailing list submissions to
> > bfl [..] ...
> >
> > To subscribe or unsubscribe via the World Wide Web, visit
> > http://lists.mech.kuleuven.be/mailman/listinfo/bfl
> > or, via email, send a message with subject or body 'help' to
> > bfl-request [..] ...
> >
> > You can reach the person managing the list at
> > bfl-owner [..] ...
> >
> > When replying, please edit your Subject line so it is more specific
> > than "Re: Contents of BFL digest..."
> >
> > Today's Topics:
> >
> > 1. Bind variables in BFL filter (Ugo Cupcic)
> >
> >
> > ---------- Forwarded message ----------
> > From: Ugo Cupcic <ugo [..] ...>
> > To: <bfl [..] ...>
> > Cc:
> > Date: Mon, 11 Aug 2014 15:45:57 +0200
> > Subject: [BFL] Bind variables in BFL filter
> > Hi,
> >
> > We're trying to use a Bayesian Filter to do some sensor fusion for
> > estimating the joints positions on a joint based robot (estimated with
> > different sensors).
> >
> > We now have two versions of the same filter (one using
> > ExtendedKalmanFilter, the other one a particle filter: BootstrapFilter)
> > that converges but haven't yet found a way of binding the BFL state
> > representation variables. In our case, we'd like to make sure that each
> > variable of the state stays within the joint limits.
> >
> > Feel free to ask for more details as I'm not exactly sure what is most
> > relevant.
> >
> >
> Hi Ugo,
>
> I am no expert, but I'll give you my opinion anyway.
> I might be wrong, but is it possible that by "binding" you actually mean
> "constraining"?
> If that is the case, than I am afraid there is no "clean" way of doing
> this.
> If the state variables are defined by Gaussian pdfs, than by definition
> those span an infinite support.
> You could try to make variables outside of the joint limits "highly
> unlikely" by filtering your measurements, and rejecting those outside a
> validation area, or carefully defining Process noise and Measurement Noise
> Covariance Matrices, but if you really want to use constraints, that
> you're actually solving an Estimation + Constrained Optimization problem,
> and you'll need different tools for that.
>
> Cheers,
>
> Enrico
>
> Cheers,
> >
> > Ugo
> >
> > PS: sorry for cross-posting on ROS-answers
> > http://answers.ros.org/question/189700/bfl-bind-variables/
> > --
> > Ugo Cupcic *Senior Software Engineer*+44 20 7700 2487ugo [..] ...
> > www.shadowrobot.com
> >
> > --
> >
> > ------------------------------
> > *Shadow Robot Company Ltd.*
> > 251 Liverpool Road, N1 1LX, UK
> > Registered Number 3308007 (England & Wales)
> >
> > RoNeX - Building Robots with ROS Made Easy
> > http://www.shadowrobot.com/products/ronex/
> > ------------------------------
> >
> >
> >
> > _______________________________________________
> > BFL mailing list
> > BFL [..] ...
> > http://lists.mech.kuleuven.be/mailman/listinfo/bfl
> >
> >
> _______________________________________________
> I hereby promise not to top-post on the
> BFL mailing list
> BFL [..] ...
> http://lists.mech.kuleuven.be/mailman/listinfo/bfl
>