Driving a model with marker coordinates

Hello everyone,

I have mocap data (x,y,z coordinates of 15 markers on pelvis and legs)
to simulate a carving swing in AnyBody. I have 100 timesteps and the
swing takes about 4 seconds in real time. I took the Gait3D model to
change it up to make the model drive a carving swing. it is actually
wotking until timestep 16. he always loses his markers (especially the
pelvis markers) and results in an kinematic error.
I have also splitted the swing up in two parts. For example: I started
the swing in timestep 44. For this timestep on it is working the next
40 steps but then he also loses his marker.

the data are quite accurate and have no significant variances. so they
are completely okay! what can I do?? whats wrong? please help me!

best regards,
tom

Hi Tom

It is a difficult question to answer without having seen the model.

I think I would look at the kinematic errors and find the largest one; this
sometimes gives you an idea on what has gone wrong.

Then you could try to exclude this marker driver and replace it with a joint
driver, basically the first thing you need to find out is which marker
driver are causing the problem.

Once you have tracked down the problem you need to change that driver, but
to be able to do so it is important to know the exact problem.

These could be:

1 The legs are too short/long try just to change the length slightly and
detect if it helps.

2 Make sure you know the function of AnyKinLinear and the Ref setting. It is
very important for these kinds of models that they are driven using the
local coordinate system on the bone as the reference.

3 Look at all you marker trajectories and make sure none of them have jumps
that may cause the kinematic problem.

I hope some of these recommendations will be useful, it is difficult to be
more specific without having seen the model.

Have a nice weekend!

Best regards

Søren, AnyBody Support


From: anyscript@yahoogroups.com [mailto:anyscript@yahoogroups.com] On Behalf
Of kollair12345
Sent: 20 November 2008 13:06
To: anyscript@yahoogroups.com
Subject: [AnyScript] Driving a model with marker coordinates

Hello everyone,

I have mocap data (x,y,z coordinates of 15 markers on pelvis and legs)
to simulate a carving swing in AnyBody. I have 100 timesteps and the
swing takes about 4 seconds in real time. I took the Gait3D model to
change it up to make the model drive a carving swing. it is actually
wotking until timestep 16. he always loses his markers (especially the
pelvis markers) and results in an kinematic error.
I have also splitted the swing up in two parts. For example: I started
the swing in timestep 44. For this timestep on it is working the next
40 steps but then he also loses his marker.

the data are quite accurate and have no significant variances. so they
are completely okay! what can I do?? whats wrong? please help me!

best regards,
tom

[Non-text portions of this message have been removed]

Hello Soren,

Thank you for the quick answer and the quite helpful recommendations.

If it is helpful I also can send you the model for better analysis of
the problem.

How can I find the largest error? Is it just the error in the output
window that overruns the error tolerance?

to 1.) I have already changed the AnthroData to fit the marker
positions as good as possible. Furthermore this helped to solve more
time steps.

to 2.) That was new for me and so I played around with it and
achieved a really solution. No the skier drives up to time step 49!!!
Really good!! Thanks a lot! Unfortunately I need to solve 100 time
steps so there is still a problem. I just changed the Ref from the
PelvisToSacrumMarkerLin from 0 to 2. If I change one of the others
(knee, foot, etc.) the error comes much earlier then before.
I have actually not much experience with the AnyKinLinear command. I
read a lot about it but I still don’t know what this is exactly for?
Same problem with the MeasureOrganizer.

to 3.) What is the biggest jump of marker trajectory that AnyBody can
solve? I checked the data several times and found no significant
variances.

Thanks a lot for you help! Do you have any further ideas to solve
this problem? As I mentioned before, the second recommendation was
really helpful!

Best regards,
Tom

— In anyscript@yahoogroups.com, AnyScript Support
<anyscriptsup@…> wrote:
>
> Hi Tom
>
>
>
> It is a difficult question to answer without having seen the model.
>
>
>
> I think I would look at the kinematic errors and find the largest
one; this
> sometimes gives you an idea on what has gone wrong.
>
>
>
> Then you could try to exclude this marker driver and replace it
with a joint
> driver, basically the first thing you need to find out is which
marker
> driver are causing the problem.
>
>
>
> Once you have tracked down the problem you need to change that
driver, but
> to be able to do so it is important to know the exact problem.
>
>
>
> These could be:
>
> 1 The legs are too short/long try just to change the length
slightly and
> detect if it helps.
>
> 2 Make sure you know the function of AnyKinLinear and the Ref
setting. It is
> very important for these kinds of models that they are driven using
the
> local coordinate system on the bone as the reference.
>
> 3 Look at all you marker trajectories and make sure none of them
have jumps
> that may cause the kinematic problem.
>
>
>
> I hope some of these recommendations will be useful, it is
difficult to be
> more specific without having seen the model.
>
>
>
> Have a nice weekend!
>
>
>
> Best regards
>
> Søren, AnyBody Support
>
>
>
> _____
>
> From: anyscript@yahoogroups.com [mailto:anyscript@yahoogroups.com]
On Behalf
> Of kollair12345
> Sent: 20 November 2008 13:06
> To: anyscript@yahoogroups.com
> Subject: [AnyScript] Driving a model with marker coordinates
>
>
>
> Hello everyone,
>
> I have mocap data (x,y,z coordinates of 15 markers on pelvis and
legs)
> to simulate a carving swing in AnyBody. I have 100 timesteps and
the
> swing takes about 4 seconds in real time. I took the Gait3D model
to
> change it up to make the model drive a carving swing. it is
actually
> wotking until timestep 16. he always loses his markers (especially
the
> pelvis markers) and results in an kinematic error.
> I have also splitted the swing up in two parts. For example: I
started
> the swing in timestep 44. For this timestep on it is working the
next
> 40 steps but then he also loses his marker.
>
> the data are quite accurate and have no significant variances. so
they
> are completely okay! what can I do?? whats wrong? please help me!
>
> best regards,
> tom
>
>
>
>
>
> [Non-text portions of this message have been removed]
>

Hi Tom

Looking for the largest error is no guarantee that it will guide you to the
root of the problem but it may do so.
The way to find the largest error is simply to look at the error message and
spot the highest number in the list of violated kinematic constraints.

Q1: ok

Q2: It is essential that you know exactly how the AnyKinLinear and the
measure organizer works when you try to drive a model like this otherwise I
am afraid it will not work.

Please have a look at this tutorial it shows how to drive a model using
markers http://www.anybodytech.com/677.0.html, it gives examples on both how
to use the measure organizer and the AnyKinLinear, if you have any questions
to this tutorial, please ask.

In short the AnyKinLinear is measuring the position between to coordinate
systems. It can do this in different ways by measuring using the global
coordinate system as basis or using one of the two coordinate system listed
in the measure. Shifting between these coordinate systems is done using the
Ref variable. If set the Ref variable so that the coordinate system of the
bone is used, then imagine that you attach your eye to this coordinate
system of the bone and follow it, this is how the measure works then.

When creating a model like this, it is normally a good idea to create a
small sketch figure with arrows showing which directions the markers are
driving; it may help visualizing the mechanics of problem. For each of the
measures visualize the coordinate system by inserting an AnyDrawRef frame.
It is also essential that you are aware of the dof of the leg.
Hip=spherical
Knee = revolute
Ankle=universal

The AnyKinLinear and similarly for the AnyKinRotational measure three
components x,y,z, for model like this you do not necessarily drive all three
of them because this will make the model over determined so you have to pick
for example to drive only the x or z direction this is what the measure
organizer is being used for.

MeasureOrganizer={0}; will pick only the x direction and so on…

Q3: Unfortunately there is no clear answer on this one, it entirely depends
on how the model is driven.

Hope this helps, otherwise please ask again.

Best regards
Søren, AnyBody Support


From: anyscript@yahoogroups.com [mailto:anyscript@yahoogroups.com] On Behalf
Of kollair12345
Sent: 25 November 2008 13:26
To: anyscript@yahoogroups.com
Subject: [AnyScript] Re: Driving a model with marker coordinates

Hello Soren,

Thank you for the quick answer and the quite helpful recommendations.

If it is helpful I also can send you the model for better analysis of
the problem.

How can I find the largest error? Is it just the error in the output
window that overruns the error tolerance?

to 1.) I have already changed the AnthroData to fit the marker
positions as good as possible. Furthermore this helped to solve more
time steps.

to 2.) That was new for me and so I played around with it and
achieved a really solution. No the skier drives up to time step 49!!!
Really good!! Thanks a lot! Unfortunately I need to solve 100 time
steps so there is still a problem. I just changed the Ref from the
PelvisToSacrumMarkerLin from 0 to 2. If I change one of the others
(knee, foot, etc.) the error comes much earlier then before.
I have actually not much experience with the AnyKinLinear command. I
read a lot about it but I still don’t know what this is exactly for?
Same problem with the MeasureOrganizer.

to 3.) What is the biggest jump of marker trajectory that AnyBody can
solve? I checked the data several times and found no significant
variances.

Thanks a lot for you help! Do you have any further ideas to solve
this problem? As I mentioned before, the second recommendation was
really helpful!

Best regards,
Tom

— In anyscript@yahoogroups.com, AnyScript Support
<anyscriptsup@…> wrote:
>
> Hi Tom
>
>
>
> It is a difficult question to answer without having seen the model.
>
>
>
> I think I would look at the kinematic errors and find the largest
one; this
> sometimes gives you an idea on what has gone wrong.
>
>
>
> Then you could try to exclude this marker driver and replace it
with a joint
> driver, basically the first thing you need to find out is which
marker
> driver are causing the problem.
>
>
>
> Once you have tracked down the problem you need to change that
driver, but
> to be able to do so it is important to know the exact problem.
>
>
>
> These could be:
>
> 1 The legs are too short/long try just to change the length
slightly and
> detect if it helps.
>
> 2 Make sure you know the function of AnyKinLinear and the Ref
setting. It is
> very important for these kinds of models that they are driven using
the
> local coordinate system on the bone as the reference.
>
> 3 Look at all you marker trajectories and make sure none of them
have jumps
> that may cause the kinematic problem.
>
>
>
> I hope some of these recommendations will be useful, it is
difficult to be
> more specific without having seen the model.
>
>
>
> Have a nice weekend!
>
>
>
> Best regards
>
> Søren, AnyBody Support
>
>
>
> _____
>
> From: anyscript@yahoogroups.com [mailto:anyscript@yahoogroups.com]
On Behalf
> Of kollair12345
> Sent: 20 November 2008 13:06
> To: anyscript@yahoogroups.com
> Subject: [AnyScript] Driving a model with marker coordinates
>
>
>
> Hello everyone,
>
> I have mocap data (x,y,z coordinates of 15 markers on pelvis and
legs)
> to simulate a carving swing in AnyBody. I have 100 timesteps and
the
> swing takes about 4 seconds in real time. I took the Gait3D model
to
> change it up to make the model drive a carving swing. it is
actually
> wotking until timestep 16. he always loses his markers (especially
the
> pelvis markers) and results in an kinematic error.
> I have also splitted the swing up in two parts. For example: I
started
> the swing in timestep 44. For this timestep on it is working the
next
> 40 steps but then he also loses his marker.
>
> the data are quite accurate and have no significant variances. so
they
> are completely okay! what can I do?? whats wrong? please help me!
>
> best regards,
> tom
>
>
>
>
>
> [Non-text portions of this message have been removed]
>