Gait3D: extraordinary force magnitude

system · June 4, 2007, 11:19pm

Hi SÃ¸ren:

I am applying Gait3D using my own data. The model seems working and
the magnitude of the joint moments are also OK. But the predicted
magnitude of muscle forces are very strange. Especially, the force in
gastrocnemius reaches over 4500 N, it’s too large; wile force in
soleus is only 200-300N, too small. I have checked your sample Gait3D,
the muscle forces are OK. I have run several other data sets, and the
results are similar. It seems there is a systematic problem in my
model, but I cannot find it. Can you look at my model and give me some
idea where the problem may be?

I have up-loaded my model: “Gait3D-revised”

Thanks,

John Wu

system · June 6, 2007, 1:29pm

Hi John

I have tested your model and found a few things

1: the position of the CentrOfPressure segment seem to be too low.
This can be corrected by adding a new node in the globalref which is
located higher, and then refering to this node in the measures used
for driving this segment.

2: The ankle eversion/inversion seems a bit strange, this might
cause problems for the muscles carrying this dof. In order to see if
this is a problem, i inserted these lines in the
JointsAndDrivers.any file

AnyReacForce AnkleEversion ={ //stc added these reactions because
AnyKinMeasureOrg
&ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
AnyKinMeasureOrg
&ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
};

The lines adds a reaction force to ankle inversion/eversion dofs. It
help the activity to drop a bit, but not enough.

3: Then i switched off the muscles and looked at the moments. The
left knee moment looks a bit strange you can see by looking at the
property

“Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.dof
0.Muscle.*.Ft”

it will give you two curves one of them is the negative moment the
other one the positive moment. Normally the negative one should be
dominant in these models and it should not change sign during the
stance phase as it does in this case, so i think this is worth
investigating a bit. I have not been able to find the cause of it,
typical reasons could be kinematics of the left leg, the way the
force is applied etc.

4: The last thing i have noticed is about the scaling, in the
AnyMan.any file it seems that your subject have a weigth of 78 kg,
but looking at the segment masses of the thigh and shank they are
unchanged compared to the original Gait3D model. I think these
values should be slight higher.

I hope it helps you move on, although it do not solve the entire
problem.

Best regards
SÃ¸ren, AnyBody support

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi SÃ¸ren:
>
> I am applying Gait3D using my own data. The model seems working and
> the magnitude of the joint moments are also OK. But the predicted
> magnitude of muscle forces are very strange. Especially, the force
in
> gastrocnemius reaches over 4500 N, it’s too large; wile force in
> soleus is only 200-300N, too small. I have checked your sample
Gait3D,
> the muscle forces are OK. I have run several other data sets, and
the
> results are similar. It seems there is a systematic problem in my
> model, but I cannot find it. Can you look at my model and give me
some
> idea where the problem may be?
>
> I have up-loaded my model: “Gait3D-revised”
>
> Thanks,
>
> John Wu
>

system · June 6, 2007, 11:24pm

Hi SÃ¸ren:

Thanks for your suggestions. I have improved my model, as suggested.

The CenterOfPressure drivers have been revised. Now, I drive
CenterOfPressure only in x-y plane, and raise it to a height of “pz”,
which can ben adjusted. By doing this, the muscle force in
gastrocnemius is reduced by 10%, still too high. (See my revised
“enviroment.any” file.)
All section masses have been scaled. But this has little influence.
(see revised “man.any”)
I have checked the joint moment by using “No_Muscle” and output the
joint muscle force. The sum of plus and minus force gives the moment.
I have added a data processing unit (“Output_kinetic.any”) to output
the joint moment. Note: M and nM respresent moment and normalized
moment, respectively. When normalized to the body mass, the joint
moments are consistent with published data.
I found that the soleus muscle seems not active and has no force
output. All external forces are balanced by the gastrocnemius muscles.
I have checked by varying the force magnitude in environment.any.
Using “F=k*force(t)” with varied “k”, I am able to change the ground
reaction force input. The output of the forces of gastrocnemius are
almost proportional to those in force plate.
In “JointsAndDrivers.any”, I have tested to switch on Reaction.Type
at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On}, the soleus
forces then appear and gastrocnemius force decrease by 1/3. It seems
the problem is at ankle joints.

I appreciate it you can take time to look at my revised model, and
give me some further suggestions.

The revised model “Gait3D-rev2” has been up-loaded.

John

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…> wrote:
>
> Hi John
>
> I have tested your model and found a few things
>
> 1: the position of the CentrOfPressure segment seem to be too low.
> This can be corrected by adding a new node in the globalref which is
> located higher, and then refering to this node in the measures used
> for driving this segment.
>
> 2: The ankle eversion/inversion seems a bit strange, this might
> cause problems for the muscles carrying this dof. In order to see if
> this is a problem, i inserted these lines in the
> JointsAndDrivers.any file
>
> AnyReacForce AnkleEversion ={ //stc added these reactions because
> AnyKinMeasureOrg
> &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> AnyKinMeasureOrg
> &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> };
>
>
> The lines adds a reaction force to ankle inversion/eversion dofs. It
> help the activity to drop a bit, but not enough.
>
>
> 3: Then i switched off the muscles and looked at the moments. The
> left knee moment looks a bit strange you can see by looking at the
> property
>
> “Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.dof
> 0.Muscle.*.Ft”
>
> it will give you two curves one of them is the negative moment the
> other one the positive moment. Normally the negative one should be
> dominant in these models and it should not change sign during the
> stance phase as it does in this case, so i think this is worth
> investigating a bit. I have not been able to find the cause of it,
> typical reasons could be kinematics of the left leg, the way the
> force is applied etc.
>
> 4: The last thing i have noticed is about the scaling, in the
> AnyMan.any file it seems that your subject have a weigth of 78 kg,
> but looking at the segment masses of the thigh and shank they are
> unchanged compared to the original Gait3D model. I think these
> values should be slight higher.
>
> I hope it helps you move on, although it do not solve the entire
> problem.
>
> Best regards
> SÃ¸ren, AnyBody support
>
>
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi SÃ¸ren:
> >
> > I am applying Gait3D using my own data. The model seems working and
> > the magnitude of the joint moments are also OK. But the predicted
> > magnitude of muscle forces are very strange. Especially, the force
> in
> > gastrocnemius reaches over 4500 N, it’s too large; wile force in
> > soleus is only 200-300N, too small. I have checked your sample
> Gait3D,
> > the muscle forces are OK. I have run several other data sets, and
> the
> > results are similar. It seems there is a systematic problem in my
> > model, but I cannot find it. Can you look at my model and give me
> some
> > idea where the problem may be?
> >
> > I have up-loaded my model: “Gait3D-revised”
> >
> > Thanks,
> >
> > John Wu
> >
>

system · June 7, 2007, 1:37pm

Hi John

I have looked a bit more into the model

I think the main problem could be the kinemtatics. I tried again to
insert a reaction force on the right and left ankle
inversion/eversion and the activity dropped from 15 to 5.8. This
might be cause by the angle between the foot and the floor(tilt), it
looks like he is walking on the lateral side of the feets.

Secondly i looked at the motion of the pelvis, if you look the
motion from the front there are some very large rotations of the
pelvis, and it can be seen that there is a large error on the Asis
markers. Therefore i tried adding reaction forces to the hip
abduction and external rotatation together with the ankle inversion
moment.

AnyReacForce LeftSideJoints ={

 AnyKinMeasureOrg &ref2=

Main.Model.HumanModel.Interface.Left.AnkleEversion;
AnyKinMeasureOrg &ref5=
Main.Model.HumanModel.Interface.Left.HipAbduction;
AnyKinMeasureOrg &ref6=
Main.Model.HumanModel.Interface.Left.HipExternalRotation;

};

AnyReacForce RightSideJoints ={

 AnyKinMeasureOrg &ref2=

Main.Model.HumanModel.Interface.Right.AnkleEversion;
AnyKinMeasureOrg &ref5=
Main.Model.HumanModel.Interface.Right.HipAbduction;
AnyKinMeasureOrg &ref6=
Main.Model.HumanModel.Interface.Right.HipExternalRotation;

};

This helped the activity to drop to 1.4. I think the explanation is
that the momentarms around the hip joint becomes too small for this
amount of pelvis “tilt”.

So i think the next step will be to adjust the motion.

I looked at the moment curves and compared them to the curves which
can be found in book "Dynamics of human gait "
http://www.kiboho.co.za/GaitCD/GaitBook.pdf

On page 42 there is a curve for the moment around the hip, knee and
ankle, when i compare the knee moment with the
curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite different?

My impression is that it is the kinematics which are causing these
problems primarily. I think the scaling and the marker postions
should be adjusted until the motion gives a smaller hip “tilt” and
the foot tilt are more horizontal and aligned with the floor.

Best regards
SÃ¸ren, AnyBody support

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi SÃ¸ren:
>
> Thanks for your suggestions. I have improved my model, as
suggested.
>
> 1. The CenterOfPressure drivers have been revised. Now, I drive
> CenterOfPressure only in x-y plane, and raise it to a height
of “pz”,
> which can ben adjusted. By doing this, the muscle force in
> gastrocnemius is reduced by 10%, still too high. (See my revised
> “enviroment.any” file.)
>
> 2. All section masses have been scaled. But this has little
influence.
> (see revised “man.any”)
>
> 3. I have checked the joint moment by using “No_Muscle” and output
the
> joint muscle force. The sum of plus and minus force gives the
moment.
> I have added a data processing unit (“Output_kinetic.any”) to
output
> the joint moment. Note: M and nM respresent moment and normalized
> moment, respectively. When normalized to the body mass, the joint
> moments are consistent with published data.
>
> 4. I found that the soleus muscle seems not active and has no force
> output. All external forces are balanced by the gastrocnemius
muscles.
> I have checked by varying the force magnitude in environment.any.
> Using “F=kforce(t)" with varied “k”, I am able to change the
ground
> reaction force input. The output of the forces of gastrocnemius are
> almost proportional to those in force plate.
>
> 5. In “JointsAndDrivers.any”, I have tested to switch on
Reaction.Type
> at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On}, the soleus
> forces then appear and gastrocnemius force decrease by 1/3. It
seems
> the problem is at ankle joints.
>
> I appreciate it you can take time to look at my revised model, and
> give me some further suggestions.
>
> The revised model “Gait3D-rev2” has been up-loaded.
>
> John
>
> — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
wrote:
> >
> > Hi John
> >
> > I have tested your model and found a few things
> >
> > 1: the position of the CentrOfPressure segment seem to be too
low.
> > This can be corrected by adding a new node in the globalref
which is
> > located higher, and then refering to this node in the measures
used
> > for driving this segment.
> >
> > 2: The ankle eversion/inversion seems a bit strange, this might
> > cause problems for the muscles carrying this dof. In order to
see if
> > this is a problem, i inserted these lines in the
> > JointsAndDrivers.any file
> >
> > AnyReacForce AnkleEversion ={ //stc added these reactions
because
> > AnyKinMeasureOrg
> > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > AnyKinMeasureOrg
> > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > };
> >
> >
> > The lines adds a reaction force to ankle inversion/eversion
dofs. It
> > help the activity to drop a bit, but not enough.
> >
> >
> > 3: Then i switched off the muscles and looked at the moments.
The
> > left knee moment looks a bit strange you can see by looking at
the
> > property
> >
>
> "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
of
> > 0.Muscle..Ft”
> >
> > it will give you two curves one of them is the negative moment
the
> > other one the positive moment. Normally the negative one should
be
> > dominant in these models and it should not change sign during
the
> > stance phase as it does in this case, so i think this is worth
> > investigating a bit. I have not been able to find the cause of
it,
> > typical reasons could be kinematics of the left leg, the way the
> > force is applied etc.
> >
> > 4: The last thing i have noticed is about the scaling, in the
> > AnyMan.any file it seems that your subject have a weigth of 78
kg,
> > but looking at the segment masses of the thigh and shank they
are
> > unchanged compared to the original Gait3D model. I think these
> > values should be slight higher.
> >
> > I hope it helps you move on, although it do not solve the entire
> > problem.
> >
> > Best regards
> > SÃ¸ren, AnyBody support
> >
> >
> >
> > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > >
> > > Hi SÃ¸ren:
> > >
> > > I am applying Gait3D using my own data. The model seems
working and
> > > the magnitude of the joint moments are also OK. But the
predicted
> > > magnitude of muscle forces are very strange. Especially, the
force
> > in
> > > gastrocnemius reaches over 4500 N, it’s too large; wile force
in
> > > soleus is only 200-300N, too small. I have checked your sample
> > Gait3D,
> > > the muscle forces are OK. I have run several other data sets,
and
> > the
> > > results are similar. It seems there is a systematic problem in
my
> > > model, but I cannot find it. Can you look at my model and give
me
> > some
> > > idea where the problem may be?
> > >
> > > I have up-loaded my model: “Gait3D-revised”
> > >
> > > Thanks,
> > >
> > > John Wu
> > >
> >
>

system · June 7, 2007, 11:48pm

Hi SÃ¸ren:

Thanks for your time and efforts. I agree with you that the problem is
likely caused by the kinematics. But, it is a tough job. I have spent
a lot of time to have my model scaled and motion markers adjusted.
When it is initialized, the model looks very good. But, it runs only
20 steps and stops. What I found is that the subject in our study is
unsymmetric. It is shown in my model, if you look at it closely. The
thigh and shank lengths are different from left to right. We also
found it in our direct physical measurements. This may cause the
problem. My questions is: how to adjust the left and right thigh and
shank length separately?

I have revised the MotionMarkerDrivers, but it does not help.

I have up-loaded my up-dated model “Gait3D-rev3”. All motion markers,
which are not used to drive the model, have been removed in the model
to make it clearere to view.

Thanks,

John

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…> wrote:
>
> Hi John
>
> I have looked a bit more into the model
>
>
> I think the main problem could be the kinemtatics. I tried again to
> insert a reaction force on the right and left ankle
> inversion/eversion and the activity dropped from 15 to 5.8. This
> might be cause by the angle between the foot and the floor(tilt), it
> looks like he is walking on the lateral side of the feets.
>
> Secondly i looked at the motion of the pelvis, if you look the
> motion from the front there are some very large rotations of the
> pelvis, and it can be seen that there is a large error on the Asis
> markers. Therefore i tried adding reaction forces to the hip
> abduction and external rotatation together with the ankle inversion
> moment.
>
>
>
> AnyReacForce LeftSideJoints ={
>
> AnyKinMeasureOrg &ref2=
> Main.Model.HumanModel.Interface.Left.AnkleEversion;
> AnyKinMeasureOrg &ref5=
> Main.Model.HumanModel.Interface.Left.HipAbduction;
> AnyKinMeasureOrg &ref6=
> Main.Model.HumanModel.Interface.Left.HipExternalRotation;
>
> };
>
>
> AnyReacForce RightSideJoints ={
>
> AnyKinMeasureOrg &ref2=
> Main.Model.HumanModel.Interface.Right.AnkleEversion;
> AnyKinMeasureOrg &ref5=
> Main.Model.HumanModel.Interface.Right.HipAbduction;
> AnyKinMeasureOrg &ref6=
> Main.Model.HumanModel.Interface.Right.HipExternalRotation;
>
> };
>
> This helped the activity to drop to 1.4. I think the explanation is
> that the momentarms around the hip joint becomes too small for this
> amount of pelvis “tilt”.
>
> So i think the next step will be to adjust the motion.
>
>
> I looked at the moment curves and compared them to the curves which
> can be found in book "Dynamics of human gait "
> http://www.kiboho.co.za/GaitCD/GaitBook.pdf
>
> On page 42 there is a curve for the moment around the hip, knee and
> ankle, when i compare the knee moment with the
> curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite different?
>
> My impression is that it is the kinematics which are causing these
> problems primarily. I think the scaling and the marker postions
> should be adjusted until the motion gives a smaller hip “tilt” and
> the foot tilt are more horizontal and aligned with the floor.
>
>
> Best regards
> SÃ¸ren, AnyBody support
>
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi SÃ¸ren:
> >
> > Thanks for your suggestions. I have improved my model, as
> suggested.
> >
> > 1. The CenterOfPressure drivers have been revised. Now, I drive
> > CenterOfPressure only in x-y plane, and raise it to a height
> of “pz”,
> > which can ben adjusted. By doing this, the muscle force in
> > gastrocnemius is reduced by 10%, still too high. (See my revised
> > “enviroment.any” file.)
> >
> > 2. All section masses have been scaled. But this has little
> influence.
> > (see revised “man.any”)
> >
> > 3. I have checked the joint moment by using “No_Muscle” and output
> the
> > joint muscle force. The sum of plus and minus force gives the
> moment.
> > I have added a data processing unit (“Output_kinetic.any”) to
> output
> > the joint moment. Note: M and nM respresent moment and normalized
> > moment, respectively. When normalized to the body mass, the joint
> > moments are consistent with published data.
> >
> > 4. I found that the soleus muscle seems not active and has no force
> > output. All external forces are balanced by the gastrocnemius
> muscles.
> > I have checked by varying the force magnitude in environment.any.
> > Using “F=kforce(t)" with varied “k”, I am able to change the
> ground
> > reaction force input. The output of the forces of gastrocnemius are
> > almost proportional to those in force plate.
> >
> > 5. In “JointsAndDrivers.any”, I have tested to switch on
> Reaction.Type
> > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On}, the soleus
> > forces then appear and gastrocnemius force decrease by 1/3. It
> seems
> > the problem is at ankle joints.
> >
> > I appreciate it you can take time to look at my revised model, and
> > give me some further suggestions.
> >
> > The revised model “Gait3D-rev2” has been up-loaded.
> >
> > John
> >
> > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> wrote:
> > >
> > > Hi John
> > >
> > > I have tested your model and found a few things
> > >
> > > 1: the position of the CentrOfPressure segment seem to be too
> low.
> > > This can be corrected by adding a new node in the globalref
> which is
> > > located higher, and then refering to this node in the measures
> used
> > > for driving this segment.
> > >
> > > 2: The ankle eversion/inversion seems a bit strange, this might
> > > cause problems for the muscles carrying this dof. In order to
> see if
> > > this is a problem, i inserted these lines in the
> > > JointsAndDrivers.any file
> > >
> > > AnyReacForce AnkleEversion ={ //stc added these reactions
> because
> > > AnyKinMeasureOrg
> > > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > AnyKinMeasureOrg
> > > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > };
> > >
> > >
> > > The lines adds a reaction force to ankle inversion/eversion
> dofs. It
> > > help the activity to drop a bit, but not enough.
> > >
> > >
> > > 3: Then i switched off the muscles and looked at the moments.
> The
> > > left knee moment looks a bit strange you can see by looking at
> the
> > > property
> > >
> >
> > "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> of
> > > 0.Muscle..Ft”
> > >
> > > it will give you two curves one of them is the negative moment
> the
> > > other one the positive moment. Normally the negative one should
> be
> > > dominant in these models and it should not change sign during
> the
> > > stance phase as it does in this case, so i think this is worth
> > > investigating a bit. I have not been able to find the cause of
> it,
> > > typical reasons could be kinematics of the left leg, the way the
> > > force is applied etc.
> > >
> > > 4: The last thing i have noticed is about the scaling, in the
> > > AnyMan.any file it seems that your subject have a weigth of 78
> kg,
> > > but looking at the segment masses of the thigh and shank they
> are
> > > unchanged compared to the original Gait3D model. I think these
> > > values should be slight higher.
> > >
> > > I hope it helps you move on, although it do not solve the entire
> > > problem.
> > >
> > > Best regards
> > > SÃ¸ren, AnyBody support
> > >
> > >
> > >
> > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > >
> > > > Hi SÃ¸ren:
> > > >
> > > > I am applying Gait3D using my own data. The model seems
> working and
> > > > the magnitude of the joint moments are also OK. But the
> predicted
> > > > magnitude of muscle forces are very strange. Especially, the
> force
> > > in
> > > > gastrocnemius reaches over 4500 N, it’s too large; wile force
> in
> > > > soleus is only 200-300N, too small. I have checked your sample
> > > Gait3D,
> > > > the muscle forces are OK. I have run several other data sets,
> and
> > > the
> > > > results are similar. It seems there is a systematic problem in
> my
> > > > model, but I cannot find it. Can you look at my model and give
> me
> > > some
> > > > idea where the problem may be?
> > > >
> > > > I have up-loaded my model: “Gait3D-revised”
> > > >
> > > > Thanks,
> > > >
> > > > John Wu
> > > >
> > >
> >
>

system · June 8, 2007, 12:30am

Hi SÃ¸ren:

I have improved the model a little bit. It can now run to the end.
But, the kinematics is still a problem. The updated model is
Gait3D_rev3, which is up-loaded.

John

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi SÃ¸ren:
>
> Thanks for your time and efforts. I agree with you that the problem is
> likely caused by the kinematics. But, it is a tough job. I have spent
> a lot of time to have my model scaled and motion markers adjusted.
> When it is initialized, the model looks very good. But, it runs only
> 20 steps and stops. What I found is that the subject in our study is
> unsymmetric. It is shown in my model, if you look at it closely. The
> thigh and shank lengths are different from left to right. We also
> found it in our direct physical measurements. This may cause the
> problem. My questions is: how to adjust the left and right thigh and
> shank length separately?
>
> I have revised the MotionMarkerDrivers, but it does not help.
>
> I have up-loaded my up-dated model “Gait3D-rev3”. All motion markers,
> which are not used to drive the model, have been removed in the model
> to make it clearere to view.
>
> Thanks,
>
> John
>
>
> — In anyscript@yahoogroups.com, “AnyBody Support” <support@> wrote:
> >
> > Hi John
> >
> > I have looked a bit more into the model
> >
> >
> > I think the main problem could be the kinemtatics. I tried again to
> > insert a reaction force on the right and left ankle
> > inversion/eversion and the activity dropped from 15 to 5.8. This
> > might be cause by the angle between the foot and the floor(tilt), it
> > looks like he is walking on the lateral side of the feets.
> >
> > Secondly i looked at the motion of the pelvis, if you look the
> > motion from the front there are some very large rotations of the
> > pelvis, and it can be seen that there is a large error on the Asis
> > markers. Therefore i tried adding reaction forces to the hip
> > abduction and external rotatation together with the ankle inversion
> > moment.
> >
> >
> >
> > AnyReacForce LeftSideJoints ={
> >
> > AnyKinMeasureOrg &ref2=
> > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > AnyKinMeasureOrg &ref5=
> > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > AnyKinMeasureOrg &ref6=
> > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> >
> > };
> >
> >
> > AnyReacForce RightSideJoints ={
> >
> > AnyKinMeasureOrg &ref2=
> > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > AnyKinMeasureOrg &ref5=
> > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > AnyKinMeasureOrg &ref6=
> > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> >
> > };
> >
> > This helped the activity to drop to 1.4. I think the explanation is
> > that the momentarms around the hip joint becomes too small for this
> > amount of pelvis “tilt”.
> >
> > So i think the next step will be to adjust the motion.
> >
> >
> > I looked at the moment curves and compared them to the curves which
> > can be found in book "Dynamics of human gait "
> > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> >
> > On page 42 there is a curve for the moment around the hip, knee and
> > ankle, when i compare the knee moment with the
> > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite different?
> >
> > My impression is that it is the kinematics which are causing these
> > problems primarily. I think the scaling and the marker postions
> > should be adjusted until the motion gives a smaller hip “tilt” and
> > the foot tilt are more horizontal and aligned with the floor.
> >
> >
> > Best regards
> > SÃ¸ren, AnyBody support
> >
> >
> > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > >
> > > Hi SÃ¸ren:
> > >
> > > Thanks for your suggestions. I have improved my model, as
> > suggested.
> > >
> > > 1. The CenterOfPressure drivers have been revised. Now, I drive
> > > CenterOfPressure only in x-y plane, and raise it to a height
> > of “pz”,
> > > which can ben adjusted. By doing this, the muscle force in
> > > gastrocnemius is reduced by 10%, still too high. (See my revised
> > > “enviroment.any” file.)
> > >
> > > 2. All section masses have been scaled. But this has little
> > influence.
> > > (see revised “man.any”)
> > >
> > > 3. I have checked the joint moment by using “No_Muscle” and output
> > the
> > > joint muscle force. The sum of plus and minus force gives the
> > moment.
> > > I have added a data processing unit (“Output_kinetic.any”) to
> > output
> > > the joint moment. Note: M and nM respresent moment and normalized
> > > moment, respectively. When normalized to the body mass, the joint
> > > moments are consistent with published data.
> > >
> > > 4. I found that the soleus muscle seems not active and has no force
> > > output. All external forces are balanced by the gastrocnemius
> > muscles.
> > > I have checked by varying the force magnitude in environment.any.
> > > Using “F=kforce(t)" with varied “k”, I am able to change the
> > ground
> > > reaction force input. The output of the forces of gastrocnemius are
> > > almost proportional to those in force plate.
> > >
> > > 5. In “JointsAndDrivers.any”, I have tested to switch on
> > Reaction.Type
> > > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On}, the soleus
> > > forces then appear and gastrocnemius force decrease by 1/3. It
> > seems
> > > the problem is at ankle joints.
> > >
> > > I appreciate it you can take time to look at my revised model, and
> > > give me some further suggestions.
> > >
> > > The revised model “Gait3D-rev2” has been up-loaded.
> > >
> > > John
> > >
> > > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> > wrote:
> > > >
> > > > Hi John
> > > >
> > > > I have tested your model and found a few things
> > > >
> > > > 1: the position of the CentrOfPressure segment seem to be too
> > low.
> > > > This can be corrected by adding a new node in the globalref
> > which is
> > > > located higher, and then refering to this node in the measures
> > used
> > > > for driving this segment.
> > > >
> > > > 2: The ankle eversion/inversion seems a bit strange, this might
> > > > cause problems for the muscles carrying this dof. In order to
> > see if
> > > > this is a problem, i inserted these lines in the
> > > > JointsAndDrivers.any file
> > > >
> > > > AnyReacForce AnkleEversion ={ //stc added these reactions
> > because
> > > > AnyKinMeasureOrg
> > > > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > AnyKinMeasureOrg
> > > > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > };
> > > >
> > > >
> > > > The lines adds a reaction force to ankle inversion/eversion
> > dofs. It
> > > > help the activity to drop a bit, but not enough.
> > > >
> > > >
> > > > 3: Then i switched off the muscles and looked at the moments.
> > The
> > > > left knee moment looks a bit strange you can see by looking at
> > the
> > > > property
> > > >
> > >
> > > "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > of
> > > > 0.Muscle..Ft”
> > > >
> > > > it will give you two curves one of them is the negative moment
> > the
> > > > other one the positive moment. Normally the negative one should
> > be
> > > > dominant in these models and it should not change sign during
> > the
> > > > stance phase as it does in this case, so i think this is worth
> > > > investigating a bit. I have not been able to find the cause of
> > it,
> > > > typical reasons could be kinematics of the left leg, the way the
> > > > force is applied etc.
> > > >
> > > > 4: The last thing i have noticed is about the scaling, in the
> > > > AnyMan.any file it seems that your subject have a weigth of 78
> > kg,
> > > > but looking at the segment masses of the thigh and shank they
> > are
> > > > unchanged compared to the original Gait3D model. I think these
> > > > values should be slight higher.
> > > >
> > > > I hope it helps you move on, although it do not solve the entire
> > > > problem.
> > > >
> > > > Best regards
> > > > SÃ¸ren, AnyBody support
> > > >
> > > >
> > > >
> > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > >
> > > > > Hi SÃ¸ren:
> > > > >
> > > > > I am applying Gait3D using my own data. The model seems
> > working and
> > > > > the magnitude of the joint moments are also OK. But the
> > predicted
> > > > > magnitude of muscle forces are very strange. Especially, the
> > force
> > > > in
> > > > > gastrocnemius reaches over 4500 N, it’s too large; wile force
> > in
> > > > > soleus is only 200-300N, too small. I have checked your sample
> > > > Gait3D,
> > > > > the muscle forces are OK. I have run several other data sets,
> > and
> > > > the
> > > > > results are similar. It seems there is a systematic problem in
> > my
> > > > > model, but I cannot find it. Can you look at my model and give
> > me
> > > > some
> > > > > idea where the problem may be?
> > > > >
> > > > > I have up-loaded my model: “Gait3D-revised”
> > > > >
> > > > > Thanks,
> > > > >
> > > > > John Wu
> > > > >
> > > >
> > >
> >
>

system · June 11, 2007, 9:51am

Hi John

I have tried to optimize the kinematics using an optimizer made by
Michael Skipper Andersen, who is working on a Ph.D. project, about
kinematic analysis of over-determinate systems. Unfortunately I have
not been doing successful doing so yet. The reason is the feet;
since there are only two markers here it creates problems for the
algorithm to determine primarily the eversion/inversion dof, in a
good way. Especially the left foot is problematic because the heel,
mallous and metatarsal markers are almost lying on straight line.
This problem will persist regardless of which method we use to
impose the movement on the model. If you a have a dataset with more
markers on the forefoot it would be easier for the algorithm to
solve this problem.

You are right that improving the kinematics manually is rather
difficult and time consuming; this was why I tried using Michael’s
optimizer on your problem. If it had been successful the output
would have been the optimized marker positions, segments lengths,
and joint angles, but I am afraid that for the current data set this
is not possible, I had to abandon this strategy after having spend
quite some time to make it work.

It is possible that scaling the left and right side differently will
improve this model. This can not be done straight away and requires
some changes to be made in the BRep.

I will try list the needed changes here:

In the AnyMan you need to make a left and right side folder
obviously, containing the segment lengths for the two sides.
In the UniformScaling file you need to create a left and right
side scaling for the legs segments.
In the BodyModels directory in the specific BodyModel file you
are using you need to introduce a new folder reference, for example
named ScalingRightLeft, in the Right and Left side folders. This
reference should point at the scaling laws just created. This folder
reference should have the same name in both folders, but point at
different scalinglaws.
In the Leg3d/Seg…any file you need to point at the scaling law
defined in the bodymodel right and left folder., instead of the
existing scaling which are there today.

This should introduce different scaling to the right and left side.

I hope scaling the left and right side differently will help solve
the problem, but the inversion and eversion of the foot will
continue to be a problem.

Best regards
SÃ¸ren, AnyBody Support

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi SÃ¸ren:
>
> I have improved the model a little bit. It can now run to the end.
> But, the kinematics is still a problem. The updated model is
> Gait3D_rev3, which is up-loaded.
>
> John
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi SÃ¸ren:
> >
> > Thanks for your time and efforts. I agree with you that the
problem is
> > likely caused by the kinematics. But, it is a tough job. I have
spent
> > a lot of time to have my model scaled and motion markers
adjusted.
> > When it is initialized, the model looks very good. But, it runs
only
> > 20 steps and stops. What I found is that the subject in our
study is
> > unsymmetric. It is shown in my model, if you look at it closely.
The
> > thigh and shank lengths are different from left to right. We also
> > found it in our direct physical measurements. This may cause the
> > problem. My questions is: how to adjust the left and right thigh
and
> > shank length separately?
> >
> > I have revised the MotionMarkerDrivers, but it does not help.
> >
> > I have up-loaded my up-dated model “Gait3D-rev3”. All motion
markers,
> > which are not used to drive the model, have been removed in the
model
> > to make it clearere to view.
> >
> > Thanks,
> >
> > John
> >
> >
> > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
wrote:
> > >
> > > Hi John
> > >
> > > I have looked a bit more into the model
> > >
> > >
> > > I think the main problem could be the kinemtatics. I tried
again to
> > > insert a reaction force on the right and left ankle
> > > inversion/eversion and the activity dropped from 15 to 5.8.
This
> > > might be cause by the angle between the foot and the floor
(tilt), it
> > > looks like he is walking on the lateral side of the feets.
> > >
> > > Secondly i looked at the motion of the pelvis, if you look the
> > > motion from the front there are some very large rotations of
the
> > > pelvis, and it can be seen that there is a large error on the
Asis
> > > markers. Therefore i tried adding reaction forces to the hip
> > > abduction and external rotatation together with the ankle
inversion
> > > moment.
> > >
> > >
> > >
> > > AnyReacForce LeftSideJoints ={
> > >
> > > AnyKinMeasureOrg &ref2=
> > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > AnyKinMeasureOrg &ref5=
> > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > AnyKinMeasureOrg &ref6=
> > > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > >
> > > };
> > >
> > >
> > > AnyReacForce RightSideJoints ={
> > >
> > > AnyKinMeasureOrg &ref2=
> > > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > AnyKinMeasureOrg &ref5=
> > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > AnyKinMeasureOrg &ref6=
> > > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > >
> > > };
> > >
> > > This helped the activity to drop to 1.4. I think the
explanation is
> > > that the momentarms around the hip joint becomes too small for
this
> > > amount of pelvis “tilt”.
> > >
> > > So i think the next step will be to adjust the motion.
> > >
> > >
> > > I looked at the moment curves and compared them to the curves
which
> > > can be found in book "Dynamics of human gait "
> > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > >
> > > On page 42 there is a curve for the moment around the hip,
knee and
> > > ankle, when i compare the knee moment with the
> > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite
different?
> > >
> > > My impression is that it is the kinematics which are causing
these
> > > problems primarily. I think the scaling and the marker
postions
> > > should be adjusted until the motion gives a smaller hip “tilt”
and
> > > the foot tilt are more horizontal and aligned with the floor.
> > >
> > >
> > > Best regards
> > > SÃ¸ren, AnyBody support
> > >
> > >
> > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > >
> > > > Hi SÃ¸ren:
> > > >
> > > > Thanks for your suggestions. I have improved my model, as
> > > suggested.
> > > >
> > > > 1. The CenterOfPressure drivers have been revised. Now, I
drive
> > > > CenterOfPressure only in x-y plane, and raise it to a height
> > > of “pz”,
> > > > which can ben adjusted. By doing this, the muscle force in
> > > > gastrocnemius is reduced by 10%, still too high. (See my
revised
> > > > “enviroment.any” file.)
> > > >
> > > > 2. All section masses have been scaled. But this has little
> > > influence.
> > > > (see revised “man.any”)
> > > >
> > > > 3. I have checked the joint moment by using “No_Muscle” and
output
> > > the
> > > > joint muscle force. The sum of plus and minus force gives
the
> > > moment.
> > > > I have added a data processing unit (“Output_kinetic.any”)
to
> > > output
> > > > the joint moment. Note: M and nM respresent moment and
normalized
> > > > moment, respectively. When normalized to the body mass, the
joint
> > > > moments are consistent with published data.
> > > >
> > > > 4. I found that the soleus muscle seems not active and has
no force
> > > > output. All external forces are balanced by the
gastrocnemius
> > > muscles.
> > > > I have checked by varying the force magnitude in
environment.any.
> > > > Using “F=kforce(t)" with varied “k”, I am able to change
the
> > > ground
> > > > reaction force input. The output of the forces of
gastrocnemius are
> > > > almost proportional to those in force plate.
> > > >
> > > > 5. In “JointsAndDrivers.any”, I have tested to switch on
> > > Reaction.Type
> > > > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On}, the
soleus
> > > > forces then appear and gastrocnemius force decrease by 1/3.
It
> > > seems
> > > > the problem is at ankle joints.
> > > >
> > > > I appreciate it you can take time to look at my revised
model, and
> > > > give me some further suggestions.
> > > >
> > > > The revised model “Gait3D-rev2” has been up-loaded.
> > > >
> > > > John
> > > >
> > > > — In anyscript@yahoogroups.com, “AnyBody Support”
<support@>
> > > wrote:
> > > > >
> > > > > Hi John
> > > > >
> > > > > I have tested your model and found a few things
> > > > >
> > > > > 1: the position of the CentrOfPressure segment seem to be
too
> > > low.
> > > > > This can be corrected by adding a new node in the
globalref
> > > which is
> > > > > located higher, and then refering to this node in the
measures
> > > used
> > > > > for driving this segment.
> > > > >
> > > > > 2: The ankle eversion/inversion seems a bit strange, this
might
> > > > > cause problems for the muscles carrying this dof. In order
to
> > > see if
> > > > > this is a problem, i inserted these lines in the
> > > > > JointsAndDrivers.any file
> > > > >
> > > > > AnyReacForce AnkleEversion ={ //stc added these
reactions
> > > because
> > > > > AnyKinMeasureOrg
> > > > > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > AnyKinMeasureOrg
> > > > > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > };
> > > > >
> > > > >
> > > > > The lines adds a reaction force to ankle
inversion/eversion
> > > dofs. It
> > > > > help the activity to drop a bit, but not enough.
> > > > >
> > > > >
> > > > > 3: Then i switched off the muscles and looked at the
moments.
> > > The
> > > > > left knee moment looks a bit strange you can see by
looking at
> > > the
> > > > > property
> > > > >
> > > >
> > >
> "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > of
> > > > > 0.Muscle..Ft”
> > > > >
> > > > > it will give you two curves one of them is the negative
moment
> > > the
> > > > > other one the positive moment. Normally the negative one
should
> > > be
> > > > > dominant in these models and it should not change sign
during
> > > the
> > > > > stance phase as it does in this case, so i think this is
worth
> > > > > investigating a bit. I have not been able to find the
cause of
> > > it,
> > > > > typical reasons could be kinematics of the left leg, the
way the
> > > > > force is applied etc.
> > > > >
> > > > > 4: The last thing i have noticed is about the scaling, in
the
> > > > > AnyMan.any file it seems that your subject have a weigth
of 78
> > > kg,
> > > > > but looking at the segment masses of the thigh and shank
they
> > > are
> > > > > unchanged compared to the original Gait3D model. I think
these
> > > > > values should be slight higher.
> > > > >
> > > > > I hope it helps you move on, although it do not solve the
entire
> > > > > problem.
> > > > >
> > > > > Best regards
> > > > > SÃ¸ren, AnyBody support
> > > > >
> > > > >
> > > > >
> > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
wrote:
> > > > > >
> > > > > > Hi SÃ¸ren:
> > > > > >
> > > > > > I am applying Gait3D using my own data. The model seems
> > > working and
> > > > > > the magnitude of the joint moments are also OK. But the
> > > predicted
> > > > > > magnitude of muscle forces are very strange. Especially,
the
> > > force
> > > > > in
> > > > > > gastrocnemius reaches over 4500 N, it’s too large; wile
force
> > > in
> > > > > > soleus is only 200-300N, too small. I have checked your
sample
> > > > > Gait3D,
> > > > > > the muscle forces are OK. I have run several other data
sets,
> > > and
> > > > > the
> > > > > > results are similar. It seems there is a systematic
problem in
> > > my
> > > > > > model, but I cannot find it. Can you look at my model
and give
> > > me
> > > > > some
> > > > > > idea where the problem may be?
> > > > > >
> > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > >
> > > > > > Thanks,
> > > > > >
> > > > > > John Wu
> > > > > >
> > > > >
> > > >
> > >
> >
>

system · June 11, 2007, 4:08pm

Hi SÃ¸ren:

I believe the best way to solve the problem is the kinematics
optimization. The model should be stable, if you can solve the
over-constrained kinemtics using your new optimizer.

If the problem is related to the ankle eversion/inversion, I have a
better solution. I can drive the ankle eversion directly, while still
driving the model using all other motion markers. We have data for all
joint angles measured using Peak system. Could you try it using your
kinematics optimizer? If it does not work, please try to put ankle
eversion to zero. I have up-loaded the ankle-eversion data and the
revised JointAndDrivers.any. (Gait3D_ankle_eversion.zip)

The unsymmetric model seems too complicated. And I am not sure if my
model will work, putting so much effects to change the human model to
unsymmetric. So, I would not proceed in that direction.

Thanks for your effects,

John

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…> wrote:
>
> Hi John
>
> I have tried to optimize the kinematics using an optimizer made by
> Michael Skipper Andersen, who is working on a Ph.D. project, about
> kinematic analysis of over-determinate systems. Unfortunately I have
> not been doing successful doing so yet. The reason is the feet;
> since there are only two markers here it creates problems for the
> algorithm to determine primarily the eversion/inversion dof, in a
> good way. Especially the left foot is problematic because the heel,
> mallous and metatarsal markers are almost lying on straight line.
> This problem will persist regardless of which method we use to
> impose the movement on the model. If you a have a dataset with more
> markers on the forefoot it would be easier for the algorithm to
> solve this problem.
>
> You are right that improving the kinematics manually is rather
> difficult and time consuming; this was why I tried using Michael’s
> optimizer on your problem. If it had been successful the output
> would have been the optimized marker positions, segments lengths,
> and joint angles, but I am afraid that for the current data set this
> is not possible, I had to abandon this strategy after having spend
> quite some time to make it work.
>
> It is possible that scaling the left and right side differently will
> improve this model. This can not be done straight away and requires
> some changes to be made in the BRep.
>
> I will try list the needed changes here:
> 1. In the AnyMan you need to make a left and right side folder
> obviously, containing the segment lengths for the two sides.
> 2. In the UniformScaling file you need to create a left and right
> side scaling for the legs segments.
> 3. In the BodyModels directory in the specific BodyModel file you
> are using you need to introduce a new folder reference, for example
> named ScalingRightLeft, in the Right and Left side folders. This
> reference should point at the scaling laws just created. This folder
> reference should have the same name in both folders, but point at
> different scalinglaws.
> 4. In the Leg3d/Seg…any file you need to point at the scaling law
> defined in the bodymodel right and left folder., instead of the
> existing scaling which are there today.
>
> This should introduce different scaling to the right and left side.
>
> I hope scaling the left and right side differently will help solve
> the problem, but the inversion and eversion of the foot will
> continue to be a problem.
>
> Best regards
> SÃ¸ren, AnyBody Support
>
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi SÃ¸ren:
> >
> > I have improved the model a little bit. It can now run to the end.
> > But, the kinematics is still a problem. The updated model is
> > Gait3D_rev3, which is up-loaded.
> >
> > John
> >
> > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > >
> > > Hi SÃ¸ren:
> > >
> > > Thanks for your time and efforts. I agree with you that the
> problem is
> > > likely caused by the kinematics. But, it is a tough job. I have
> spent
> > > a lot of time to have my model scaled and motion markers
> adjusted.
> > > When it is initialized, the model looks very good. But, it runs
> only
> > > 20 steps and stops. What I found is that the subject in our
> study is
> > > unsymmetric. It is shown in my model, if you look at it closely.
> The
> > > thigh and shank lengths are different from left to right. We also
> > > found it in our direct physical measurements. This may cause the
> > > problem. My questions is: how to adjust the left and right thigh
> and
> > > shank length separately?
> > >
> > > I have revised the MotionMarkerDrivers, but it does not help.
> > >
> > > I have up-loaded my up-dated model “Gait3D-rev3”. All motion
> markers,
> > > which are not used to drive the model, have been removed in the
> model
> > > to make it clearere to view.
> > >
> > > Thanks,
> > >
> > > John
> > >
> > >
> > > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> wrote:
> > > >
> > > > Hi John
> > > >
> > > > I have looked a bit more into the model
> > > >
> > > >
> > > > I think the main problem could be the kinemtatics. I tried
> again to
> > > > insert a reaction force on the right and left ankle
> > > > inversion/eversion and the activity dropped from 15 to 5.8.
> This
> > > > might be cause by the angle between the foot and the floor
> (tilt), it
> > > > looks like he is walking on the lateral side of the feets.
> > > >
> > > > Secondly i looked at the motion of the pelvis, if you look the
> > > > motion from the front there are some very large rotations of
> the
> > > > pelvis, and it can be seen that there is a large error on the
> Asis
> > > > markers. Therefore i tried adding reaction forces to the hip
> > > > abduction and external rotatation together with the ankle
> inversion
> > > > moment.
> > > >
> > > >
> > > >
> > > > AnyReacForce LeftSideJoints ={
> > > >
> > > > AnyKinMeasureOrg &ref2=
> > > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > AnyKinMeasureOrg &ref5=
> > > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > > AnyKinMeasureOrg &ref6=
> > > > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > > >
> > > > };
> > > >
> > > >
> > > > AnyReacForce RightSideJoints ={
> > > >
> > > > AnyKinMeasureOrg &ref2=
> > > > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > AnyKinMeasureOrg &ref5=
> > > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > > AnyKinMeasureOrg &ref6=
> > > > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > > >
> > > > };
> > > >
> > > > This helped the activity to drop to 1.4. I think the
> explanation is
> > > > that the momentarms around the hip joint becomes too small for
> this
> > > > amount of pelvis “tilt”.
> > > >
> > > > So i think the next step will be to adjust the motion.
> > > >
> > > >
> > > > I looked at the moment curves and compared them to the curves
> which
> > > > can be found in book "Dynamics of human gait "
> > > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > > >
> > > > On page 42 there is a curve for the moment around the hip,
> knee and
> > > > ankle, when i compare the knee moment with the
> > > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite
> different?
> > > >
> > > > My impression is that it is the kinematics which are causing
> these
> > > > problems primarily. I think the scaling and the marker
> postions
> > > > should be adjusted until the motion gives a smaller hip “tilt”
> and
> > > > the foot tilt are more horizontal and aligned with the floor.
> > > >
> > > >
> > > > Best regards
> > > > SÃ¸ren, AnyBody support
> > > >
> > > >
> > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > >
> > > > > Hi SÃ¸ren:
> > > > >
> > > > > Thanks for your suggestions. I have improved my model, as
> > > > suggested.
> > > > >
> > > > > 1. The CenterOfPressure drivers have been revised. Now, I
> drive
> > > > > CenterOfPressure only in x-y plane, and raise it to a height
> > > > of “pz”,
> > > > > which can ben adjusted. By doing this, the muscle force in
> > > > > gastrocnemius is reduced by 10%, still too high. (See my
> revised
> > > > > “enviroment.any” file.)
> > > > >
> > > > > 2. All section masses have been scaled. But this has little
> > > > influence.
> > > > > (see revised “man.any”)
> > > > >
> > > > > 3. I have checked the joint moment by using “No_Muscle” and
> output
> > > > the
> > > > > joint muscle force. The sum of plus and minus force gives
> the
> > > > moment.
> > > > > I have added a data processing unit (“Output_kinetic.any”)
> to
> > > > output
> > > > > the joint moment. Note: M and nM respresent moment and
> normalized
> > > > > moment, respectively. When normalized to the body mass, the
> joint
> > > > > moments are consistent with published data.
> > > > >
> > > > > 4. I found that the soleus muscle seems not active and has
> no force
> > > > > output. All external forces are balanced by the
> gastrocnemius
> > > > muscles.
> > > > > I have checked by varying the force magnitude in
> environment.any.
> > > > > Using “F=kforce(t)" with varied “k”, I am able to change
> the
> > > > ground
> > > > > reaction force input. The output of the forces of
> gastrocnemius are
> > > > > almost proportional to those in force plate.
> > > > >
> > > > > 5. In “JointsAndDrivers.any”, I have tested to switch on
> > > > Reaction.Type
> > > > > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On}, the
> soleus
> > > > > forces then appear and gastrocnemius force decrease by 1/3.
> It
> > > > seems
> > > > > the problem is at ankle joints.
> > > > >
> > > > > I appreciate it you can take time to look at my revised
> model, and
> > > > > give me some further suggestions.
> > > > >
> > > > > The revised model “Gait3D-rev2” has been up-loaded.
> > > > >
> > > > > John
> > > > >
> > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> <support@>
> > > > wrote:
> > > > > >
> > > > > > Hi John
> > > > > >
> > > > > > I have tested your model and found a few things
> > > > > >
> > > > > > 1: the position of the CentrOfPressure segment seem to be
> too
> > > > low.
> > > > > > This can be corrected by adding a new node in the
> globalref
> > > > which is
> > > > > > located higher, and then refering to this node in the
> measures
> > > > used
> > > > > > for driving this segment.
> > > > > >
> > > > > > 2: The ankle eversion/inversion seems a bit strange, this
> might
> > > > > > cause problems for the muscles carrying this dof. In order
> to
> > > > see if
> > > > > > this is a problem, i inserted these lines in the
> > > > > > JointsAndDrivers.any file
> > > > > >
> > > > > > AnyReacForce AnkleEversion ={ //stc added these
> reactions
> > > > because
> > > > > > AnyKinMeasureOrg
> > > > > > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > AnyKinMeasureOrg
> > > > > > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > };
> > > > > >
> > > > > >
> > > > > > The lines adds a reaction force to ankle
> inversion/eversion
> > > > dofs. It
> > > > > > help the activity to drop a bit, but not enough.
> > > > > >
> > > > > >
> > > > > > 3: Then i switched off the muscles and looked at the
> moments.
> > > > The
> > > > > > left knee moment looks a bit strange you can see by
> looking at
> > > > the
> > > > > > property
> > > > > >
> > > > >
> > > >
> > "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > > of
> > > > > > 0.Muscle..Ft”
> > > > > >
> > > > > > it will give you two curves one of them is the negative
> moment
> > > > the
> > > > > > other one the positive moment. Normally the negative one
> should
> > > > be
> > > > > > dominant in these models and it should not change sign
> during
> > > > the
> > > > > > stance phase as it does in this case, so i think this is
> worth
> > > > > > investigating a bit. I have not been able to find the
> cause of
> > > > it,
> > > > > > typical reasons could be kinematics of the left leg, the
> way the
> > > > > > force is applied etc.
> > > > > >
> > > > > > 4: The last thing i have noticed is about the scaling, in
> the
> > > > > > AnyMan.any file it seems that your subject have a weigth
> of 78
> > > > kg,
> > > > > > but looking at the segment masses of the thigh and shank
> they
> > > > are
> > > > > > unchanged compared to the original Gait3D model. I think
> these
> > > > > > values should be slight higher.
> > > > > >
> > > > > > I hope it helps you move on, although it do not solve the
> entire
> > > > > > problem.
> > > > > >
> > > > > > Best regards
> > > > > > SÃ¸ren, AnyBody support
> > > > > >
> > > > > >
> > > > > >
> > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> wrote:
> > > > > > >
> > > > > > > Hi SÃ¸ren:
> > > > > > >
> > > > > > > I am applying Gait3D using my own data. The model seems
> > > > working and
> > > > > > > the magnitude of the joint moments are also OK. But the
> > > > predicted
> > > > > > > magnitude of muscle forces are very strange. Especially,
> the
> > > > force
> > > > > > in
> > > > > > > gastrocnemius reaches over 4500 N, it’s too large; wile
> force
> > > > in
> > > > > > > soleus is only 200-300N, too small. I have checked your
> sample
> > > > > > Gait3D,
> > > > > > > the muscle forces are OK. I have run several other data
> sets,
> > > > and
> > > > > > the
> > > > > > > results are similar. It seems there is a systematic
> problem in
> > > > my
> > > > > > > model, but I cannot find it. Can you look at my model
> and give
> > > > me
> > > > > > some
> > > > > > > idea where the problem may be?
> > > > > > >
> > > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > > >
> > > > > > > Thanks,
> > > > > > >
> > > > > > > John Wu
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

system · June 12, 2007, 9:20am

Hi John

The kinematic optimizer is working on a model which has a fixed
topology, so i can not specify a specific dof like the eversion of
the ankle to be rigid unfortunately.

You are writing that you have the joint angles measured in Peak
system, maybee it would be possible for you to drive the model using
these angles directly?. This requires that you know exactly how
these angles are measured and then create new rotational measures in
AnyBody reflecting this. If these two models match kinematically
joint angle drivers should be a good solution.

Best regards
SÃ¸ren

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi SÃ¸ren:
>
> I believe the best way to solve the problem is the kinematics
> optimization. The model should be stable, if you can solve the
> over-constrained kinemtics using your new optimizer.
>
> If the problem is related to the ankle eversion/inversion, I have a
> better solution. I can drive the ankle eversion directly, while
still
> driving the model using all other motion markers. We have data for
all
> joint angles measured using Peak system. Could you try it using
your
> kinematics optimizer? If it does not work, please try to put ankle
> eversion to zero. I have up-loaded the ankle-eversion data and the
> revised JointAndDrivers.any. (Gait3D_ankle_eversion.zip)
>
> The unsymmetric model seems too complicated. And I am not sure if
my
> model will work, putting so much effects to change the human model
to
> unsymmetric. So, I would not proceed in that direction.
>
> Thanks for your effects,
>
> John
>
> — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
wrote:
> >
> > Hi John
> >
> > I have tried to optimize the kinematics using an optimizer made
by
> > Michael Skipper Andersen, who is working on a Ph.D. project,
about
> > kinematic analysis of over-determinate systems. Unfortunately I
have
> > not been doing successful doing so yet. The reason is the feet;
> > since there are only two markers here it creates problems for
the
> > algorithm to determine primarily the eversion/inversion dof, in
a
> > good way. Especially the left foot is problematic because the
heel,
> > mallous and metatarsal markers are almost lying on straight
line.
> > This problem will persist regardless of which method we use to
> > impose the movement on the model. If you a have a dataset with
more
> > markers on the forefoot it would be easier for the algorithm to
> > solve this problem.
> >
> > You are right that improving the kinematics manually is rather
> > difficult and time consuming; this was why I tried using
Michael’s
> > optimizer on your problem. If it had been successful the output
> > would have been the optimized marker positions, segments
lengths,
> > and joint angles, but I am afraid that for the current data set
this
> > is not possible, I had to abandon this strategy after having
spend
> > quite some time to make it work.
> >
> > It is possible that scaling the left and right side differently
will
> > improve this model. This can not be done straight away and
requires
> > some changes to be made in the BRep.
> >
> > I will try list the needed changes here:
> > 1. In the AnyMan you need to make a left and right side folder
> > obviously, containing the segment lengths for the two sides.
> > 2. In the UniformScaling file you need to create a left and
right
> > side scaling for the legs segments.
> > 3. In the BodyModels directory in the specific BodyModel file
you
> > are using you need to introduce a new folder reference, for
example
> > named ScalingRightLeft, in the Right and Left side folders. This
> > reference should point at the scaling laws just created. This
folder
> > reference should have the same name in both folders, but point
at
> > different scalinglaws.
> > 4. In the Leg3d/Seg…any file you need to point at the scaling
law
> > defined in the bodymodel right and left folder., instead of the
> > existing scaling which are there today.
> >
> > This should introduce different scaling to the right and left
side.
> >
> > I hope scaling the left and right side differently will help
solve
> > the problem, but the inversion and eversion of the foot will
> > continue to be a problem.
> >
> > Best regards
> > SÃ¸ren, AnyBody Support
> >
> >
> > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > >
> > > Hi SÃ¸ren:
> > >
> > > I have improved the model a little bit. It can now run to the
end.
> > > But, the kinematics is still a problem. The updated model is
> > > Gait3D_rev3, which is up-loaded.
> > >
> > > John
> > >
> > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > >
> > > > Hi SÃ¸ren:
> > > >
> > > > Thanks for your time and efforts. I agree with you that the
> > problem is
> > > > likely caused by the kinematics. But, it is a tough job. I
have
> > spent
> > > > a lot of time to have my model scaled and motion markers
> > adjusted.
> > > > When it is initialized, the model looks very good. But, it
runs
> > only
> > > > 20 steps and stops. What I found is that the subject in our
> > study is
> > > > unsymmetric. It is shown in my model, if you look at it
closely.
> > The
> > > > thigh and shank lengths are different from left to right. We
also
> > > > found it in our direct physical measurements. This may cause
the
> > > > problem. My questions is: how to adjust the left and right
thigh
> > and
> > > > shank length separately?
> > > >
> > > > I have revised the MotionMarkerDrivers, but it does not help.
> > > >
> > > > I have up-loaded my up-dated model “Gait3D-rev3”. All motion
> > markers,
> > > > which are not used to drive the model, have been removed in
the
> > model
> > > > to make it clearere to view.
> > > >
> > > > Thanks,
> > > >
> > > > John
> > > >
> > > >
> > > > — In anyscript@yahoogroups.com, “AnyBody Support”
<support@>
> > wrote:
> > > > >
> > > > > Hi John
> > > > >
> > > > > I have looked a bit more into the model
> > > > >
> > > > >
> > > > > I think the main problem could be the kinemtatics. I
tried
> > again to
> > > > > insert a reaction force on the right and left ankle
> > > > > inversion/eversion and the activity dropped from 15 to
5.8.
> > This
> > > > > might be cause by the angle between the foot and the floor
> > (tilt), it
> > > > > looks like he is walking on the lateral side of the feets.
> > > > >
> > > > > Secondly i looked at the motion of the pelvis, if you look
the
> > > > > motion from the front there are some very large rotations
of
> > the
> > > > > pelvis, and it can be seen that there is a large error on
the
> > Asis
> > > > > markers. Therefore i tried adding reaction forces to the
hip
> > > > > abduction and external rotatation together with the ankle
> > inversion
> > > > > moment.
> > > > >
> > > > >
> > > > >
> > > > > AnyReacForce LeftSideJoints ={
> > > > >
> > > > > AnyKinMeasureOrg &ref2=
> > > > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > AnyKinMeasureOrg &ref5=
> > > > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > > > AnyKinMeasureOrg &ref6=
> > > > > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > > > >
> > > > > };
> > > > >
> > > > >
> > > > > AnyReacForce RightSideJoints ={
> > > > >
> > > > > AnyKinMeasureOrg &ref2=
> > > > > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > AnyKinMeasureOrg &ref5=
> > > > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > > > AnyKinMeasureOrg &ref6=
> > > > > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > > > >
> > > > > };
> > > > >
> > > > > This helped the activity to drop to 1.4. I think the
> > explanation is
> > > > > that the momentarms around the hip joint becomes too small
for
> > this
> > > > > amount of pelvis “tilt”.
> > > > >
> > > > > So i think the next step will be to adjust the motion.
> > > > >
> > > > >
> > > > > I looked at the moment curves and compared them to the
curves
> > which
> > > > > can be found in book "Dynamics of human gait "
> > > > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > > > >
> > > > > On page 42 there is a curve for the moment around the hip,
> > knee and
> > > > > ankle, when i compare the knee moment with the
> > > > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite
> > different?
> > > > >
> > > > > My impression is that it is the kinematics which are
causing
> > these
> > > > > problems primarily. I think the scaling and the marker
> > postions
> > > > > should be adjusted until the motion gives a smaller
hip “tilt”
> > and
> > > > > the foot tilt are more horizontal and aligned with the
floor.
> > > > >
> > > > >
> > > > > Best regards
> > > > > SÃ¸ren, AnyBody support
> > > > >
> > > > >
> > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
wrote:
> > > > > >
> > > > > > Hi SÃ¸ren:
> > > > > >
> > > > > > Thanks for your suggestions. I have improved my model,
as
> > > > > suggested.
> > > > > >
> > > > > > 1. The CenterOfPressure drivers have been revised. Now,
I
> > drive
> > > > > > CenterOfPressure only in x-y plane, and raise it to a
height
> > > > > of “pz”,
> > > > > > which can ben adjusted. By doing this, the muscle force
in
> > > > > > gastrocnemius is reduced by 10%, still too high. (See my
> > revised
> > > > > > “enviroment.any” file.)
> > > > > >
> > > > > > 2. All section masses have been scaled. But this has
little
> > > > > influence.
> > > > > > (see revised “man.any”)
> > > > > >
> > > > > > 3. I have checked the joint moment by using “No_Muscle”
and
> > output
> > > > > the
> > > > > > joint muscle force. The sum of plus and minus force
gives
> > the
> > > > > moment.
> > > > > > I have added a data processing unit
(“Output_kinetic.any”)
> > to
> > > > > output
> > > > > > the joint moment. Note: M and nM respresent moment and
> > normalized
> > > > > > moment, respectively. When normalized to the body mass,
the
> > joint
> > > > > > moments are consistent with published data.
> > > > > >
> > > > > > 4. I found that the soleus muscle seems not active and
has
> > no force
> > > > > > output. All external forces are balanced by the
> > gastrocnemius
> > > > > muscles.
> > > > > > I have checked by varying the force magnitude in
> > environment.any.
> > > > > > Using “F=kforce(t)" with varied “k”, I am able to
change
> > the
> > > > > ground
> > > > > > reaction force input. The output of the forces of
> > gastrocnemius are
> > > > > > almost proportional to those in force plate.
> > > > > >
> > > > > > 5. In “JointsAndDrivers.any”, I have tested to switch on
> > > > > Reaction.Type
> > > > > > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On},
the
> > soleus
> > > > > > forces then appear and gastrocnemius force decrease by
1/3.
> > It
> > > > > seems
> > > > > > the problem is at ankle joints.
> > > > > >
> > > > > > I appreciate it you can take time to look at my revised
> > model, and
> > > > > > give me some further suggestions.
> > > > > >
> > > > > > The revised model “Gait3D-rev2” has been up-loaded.
> > > > > >
> > > > > > John
> > > > > >
> > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > <support@>
> > > > > wrote:
> > > > > > >
> > > > > > > Hi John
> > > > > > >
> > > > > > > I have tested your model and found a few things
> > > > > > >
> > > > > > > 1: the position of the CentrOfPressure segment seem to
be
> > too
> > > > > low.
> > > > > > > This can be corrected by adding a new node in the
> > globalref
> > > > > which is
> > > > > > > located higher, and then refering to this node in the
> > measures
> > > > > used
> > > > > > > for driving this segment.
> > > > > > >
> > > > > > > 2: The ankle eversion/inversion seems a bit strange,
this
> > might
> > > > > > > cause problems for the muscles carrying this dof. In
order
> > to
> > > > > see if
> > > > > > > this is a problem, i inserted these lines in the
> > > > > > > JointsAndDrivers.any file
> > > > > > >
> > > > > > > AnyReacForce AnkleEversion ={ //stc added these
> > reactions
> > > > > because
> > > > > > > AnyKinMeasureOrg
> > > > > > >
&ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > AnyKinMeasureOrg
> > > > > > >
&ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > };
> > > > > > >
> > > > > > >
> > > > > > > The lines adds a reaction force to ankle
> > inversion/eversion
> > > > > dofs. It
> > > > > > > help the activity to drop a bit, but not enough.
> > > > > > >
> > > > > > >
> > > > > > > 3: Then i switched off the muscles and looked at the
> > moments.
> > > > > The
> > > > > > > left knee moment looks a bit strange you can see by
> > looking at
> > > > > the
> > > > > > > property
> > > > > > >
> > > > > >
> > > > >
> >
> "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > > > of
> > > > > > > 0.Muscle..Ft”
> > > > > > >
> > > > > > > it will give you two curves one of them is the
negative
> > moment
> > > > > the
> > > > > > > other one the positive moment. Normally the negative
one
> > should
> > > > > be
> > > > > > > dominant in these models and it should not change sign
> > during
> > > > > the
> > > > > > > stance phase as it does in this case, so i think this
is
> > worth
> > > > > > > investigating a bit. I have not been able to find the
> > cause of
> > > > > it,
> > > > > > > typical reasons could be kinematics of the left leg,
the
> > way the
> > > > > > > force is applied etc.
> > > > > > >
> > > > > > > 4: The last thing i have noticed is about the scaling,
in
> > the
> > > > > > > AnyMan.any file it seems that your subject have a
weigth
> > of 78
> > > > > kg,
> > > > > > > but looking at the segment masses of the thigh and
shank
> > they
> > > > > are
> > > > > > > unchanged compared to the original Gait3D model. I
think
> > these
> > > > > > > values should be slight higher.
> > > > > > >
> > > > > > > I hope it helps you move on, although it do not solve
the
> > entire
> > > > > > > problem.
> > > > > > >
> > > > > > > Best regards
> > > > > > > SÃ¸ren, AnyBody support
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> > wrote:
> > > > > > > >
> > > > > > > > Hi SÃ¸ren:
> > > > > > > >
> > > > > > > > I am applying Gait3D using my own data. The model
seems
> > > > > working and
> > > > > > > > the magnitude of the joint moments are also OK. But
the
> > > > > predicted
> > > > > > > > magnitude of muscle forces are very strange.
Especially,
> > the
> > > > > force
> > > > > > > in
> > > > > > > > gastrocnemius reaches over 4500 N, it’s too large;
wile
> > force
> > > > > in
> > > > > > > > soleus is only 200-300N, too small. I have checked
your
> > sample
> > > > > > > Gait3D,
> > > > > > > > the muscle forces are OK. I have run several other
data
> > sets,
> > > > > and
> > > > > > > the
> > > > > > > > results are similar. It seems there is a systematic
> > problem in
> > > > > my
> > > > > > > > model, but I cannot find it. Can you look at my
model
> > and give
> > > > > me
> > > > > > > some
> > > > > > > > idea where the problem may be?
> > > > > > > >
> > > > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > > > >
> > > > > > > > Thanks,
> > > > > > > >
> > > > > > > > John Wu
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

system · June 13, 2007, 4:36am

Hi Soeren:

Finally I got my model working. It’s now a “clean” solution, without
using AnkleEversion driver, just driving via motion markers. It seems
the JointAndDrivers.any in Gait3D does not work on my data. I have
revised the driver at three driving ponts, and it works well now. The
gait seems normal. I am still working on it to test my another data
sets. I have four sets of gait data collected from four different
subjects. I will let you know the results and what I did.

Thanks for your help,

John

— In anyscript@yahoogroups.com, “AnyBody Support” <support@…> wrote:
>
> Hi John
>
> The kinematic optimizer is working on a model which has a fixed
> topology, so i can not specify a specific dof like the eversion of
> the ankle to be rigid unfortunately.
>
> You are writing that you have the joint angles measured in Peak
> system, maybee it would be possible for you to drive the model using
> these angles directly?. This requires that you know exactly how
> these angles are measured and then create new rotational measures in
> AnyBody reflecting this. If these two models match kinematically
> joint angle drivers should be a good solution.
>
> Best regards
> SÃ¸ren
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi SÃ¸ren:
> >
> > I believe the best way to solve the problem is the kinematics
> > optimization. The model should be stable, if you can solve the
> > over-constrained kinemtics using your new optimizer.
> >
> > If the problem is related to the ankle eversion/inversion, I have a
> > better solution. I can drive the ankle eversion directly, while
> still
> > driving the model using all other motion markers. We have data for
> all
> > joint angles measured using Peak system. Could you try it using
> your
> > kinematics optimizer? If it does not work, please try to put ankle
> > eversion to zero. I have up-loaded the ankle-eversion data and the
> > revised JointAndDrivers.any. (Gait3D_ankle_eversion.zip)
> >
> > The unsymmetric model seems too complicated. And I am not sure if
> my
> > model will work, putting so much effects to change the human model
> to
> > unsymmetric. So, I would not proceed in that direction.
> >
> > Thanks for your effects,
> >
> > John
> >
> > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> wrote:
> > >
> > > Hi John
> > >
> > > I have tried to optimize the kinematics using an optimizer made
> by
> > > Michael Skipper Andersen, who is working on a Ph.D. project,
> about
> > > kinematic analysis of over-determinate systems. Unfortunately I
> have
> > > not been doing successful doing so yet. The reason is the feet;
> > > since there are only two markers here it creates problems for
> the
> > > algorithm to determine primarily the eversion/inversion dof, in
> a
> > > good way. Especially the left foot is problematic because the
> heel,
> > > mallous and metatarsal markers are almost lying on straight
> line.
> > > This problem will persist regardless of which method we use to
> > > impose the movement on the model. If you a have a dataset with
> more
> > > markers on the forefoot it would be easier for the algorithm to
> > > solve this problem.
> > >
> > > You are right that improving the kinematics manually is rather
> > > difficult and time consuming; this was why I tried using
> Michael’s
> > > optimizer on your problem. If it had been successful the output
> > > would have been the optimized marker positions, segments
> lengths,
> > > and joint angles, but I am afraid that for the current data set
> this
> > > is not possible, I had to abandon this strategy after having
> spend
> > > quite some time to make it work.
> > >
> > > It is possible that scaling the left and right side differently
> will
> > > improve this model. This can not be done straight away and
> requires
> > > some changes to be made in the BRep.
> > >
> > > I will try list the needed changes here:
> > > 1. In the AnyMan you need to make a left and right side folder
> > > obviously, containing the segment lengths for the two sides.
> > > 2. In the UniformScaling file you need to create a left and
> right
> > > side scaling for the legs segments.
> > > 3. In the BodyModels directory in the specific BodyModel file
> you
> > > are using you need to introduce a new folder reference, for
> example
> > > named ScalingRightLeft, in the Right and Left side folders. This
> > > reference should point at the scaling laws just created. This
> folder
> > > reference should have the same name in both folders, but point
> at
> > > different scalinglaws.
> > > 4. In the Leg3d/Seg…any file you need to point at the scaling
> law
> > > defined in the bodymodel right and left folder., instead of the
> > > existing scaling which are there today.
> > >
> > > This should introduce different scaling to the right and left
> side.
> > >
> > > I hope scaling the left and right side differently will help
> solve
> > > the problem, but the inversion and eversion of the foot will
> > > continue to be a problem.
> > >
> > > Best regards
> > > SÃ¸ren, AnyBody Support
> > >
> > >
> > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > >
> > > > Hi SÃ¸ren:
> > > >
> > > > I have improved the model a little bit. It can now run to the
> end.
> > > > But, the kinematics is still a problem. The updated model is
> > > > Gait3D_rev3, which is up-loaded.
> > > >
> > > > John
> > > >
> > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > >
> > > > > Hi SÃ¸ren:
> > > > >
> > > > > Thanks for your time and efforts. I agree with you that the
> > > problem is
> > > > > likely caused by the kinematics. But, it is a tough job. I
> have
> > > spent
> > > > > a lot of time to have my model scaled and motion markers
> > > adjusted.
> > > > > When it is initialized, the model looks very good. But, it
> runs
> > > only
> > > > > 20 steps and stops. What I found is that the subject in our
> > > study is
> > > > > unsymmetric. It is shown in my model, if you look at it
> closely.
> > > The
> > > > > thigh and shank lengths are different from left to right. We
> also
> > > > > found it in our direct physical measurements. This may cause
> the
> > > > > problem. My questions is: how to adjust the left and right
> thigh
> > > and
> > > > > shank length separately?
> > > > >
> > > > > I have revised the MotionMarkerDrivers, but it does not help.
> > > > >
> > > > > I have up-loaded my up-dated model “Gait3D-rev3”. All motion
> > > markers,
> > > > > which are not used to drive the model, have been removed in
> the
> > > model
> > > > > to make it clearere to view.
> > > > >
> > > > > Thanks,
> > > > >
> > > > > John
> > > > >
> > > > >
> > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> <support@>
> > > wrote:
> > > > > >
> > > > > > Hi John
> > > > > >
> > > > > > I have looked a bit more into the model
> > > > > >
> > > > > >
> > > > > > I think the main problem could be the kinemtatics. I
> tried
> > > again to
> > > > > > insert a reaction force on the right and left ankle
> > > > > > inversion/eversion and the activity dropped from 15 to
> 5.8.
> > > This
> > > > > > might be cause by the angle between the foot and the floor
> > > (tilt), it
> > > > > > looks like he is walking on the lateral side of the feets.
> > > > > >
> > > > > > Secondly i looked at the motion of the pelvis, if you look
> the
> > > > > > motion from the front there are some very large rotations
> of
> > > the
> > > > > > pelvis, and it can be seen that there is a large error on
> the
> > > Asis
> > > > > > markers. Therefore i tried adding reaction forces to the
> hip
> > > > > > abduction and external rotatation together with the ankle
> > > inversion
> > > > > > moment.
> > > > > >
> > > > > >
> > > > > >
> > > > > > AnyReacForce LeftSideJoints ={
> > > > > >
> > > > > > AnyKinMeasureOrg &ref2=
> > > > > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > AnyKinMeasureOrg &ref5=
> > > > > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > > > > AnyKinMeasureOrg &ref6=
> > > > > > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > > > > >
> > > > > > };
> > > > > >
> > > > > >
> > > > > > AnyReacForce RightSideJoints ={
> > > > > >
> > > > > > AnyKinMeasureOrg &ref2=
> > > > > > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > AnyKinMeasureOrg &ref5=
> > > > > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > > > > AnyKinMeasureOrg &ref6=
> > > > > > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > > > > >
> > > > > > };
> > > > > >
> > > > > > This helped the activity to drop to 1.4. I think the
> > > explanation is
> > > > > > that the momentarms around the hip joint becomes too small
> for
> > > this
> > > > > > amount of pelvis “tilt”.
> > > > > >
> > > > > > So i think the next step will be to adjust the motion.
> > > > > >
> > > > > >
> > > > > > I looked at the moment curves and compared them to the
> curves
> > > which
> > > > > > can be found in book "Dynamics of human gait "
> > > > > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > > > > >
> > > > > > On page 42 there is a curve for the moment around the hip,
> > > knee and
> > > > > > ankle, when i compare the knee moment with the
> > > > > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite
> > > different?
> > > > > >
> > > > > > My impression is that it is the kinematics which are
> causing
> > > these
> > > > > > problems primarily. I think the scaling and the marker
> > > postions
> > > > > > should be adjusted until the motion gives a smaller
> hip “tilt”
> > > and
> > > > > > the foot tilt are more horizontal and aligned with the
> floor.
> > > > > >
> > > > > >
> > > > > > Best regards
> > > > > > SÃ¸ren, AnyBody support
> > > > > >
> > > > > >
> > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> wrote:
> > > > > > >
> > > > > > > Hi SÃ¸ren:
> > > > > > >
> > > > > > > Thanks for your suggestions. I have improved my model,
> as
> > > > > > suggested.
> > > > > > >
> > > > > > > 1. The CenterOfPressure drivers have been revised. Now,
> I
> > > drive
> > > > > > > CenterOfPressure only in x-y plane, and raise it to a
> height
> > > > > > of “pz”,
> > > > > > > which can ben adjusted. By doing this, the muscle force
> in
> > > > > > > gastrocnemius is reduced by 10%, still too high. (See my
> > > revised
> > > > > > > “enviroment.any” file.)
> > > > > > >
> > > > > > > 2. All section masses have been scaled. But this has
> little
> > > > > > influence.
> > > > > > > (see revised “man.any”)
> > > > > > >
> > > > > > > 3. I have checked the joint moment by using “No_Muscle”
> and
> > > output
> > > > > > the
> > > > > > > joint muscle force. The sum of plus and minus force
> gives
> > > the
> > > > > > moment.
> > > > > > > I have added a data processing unit
> (“Output_kinetic.any”)
> > > to
> > > > > > output
> > > > > > > the joint moment. Note: M and nM respresent moment and
> > > normalized
> > > > > > > moment, respectively. When normalized to the body mass,
> the
> > > joint
> > > > > > > moments are consistent with published data.
> > > > > > >
> > > > > > > 4. I found that the soleus muscle seems not active and
> has
> > > no force
> > > > > > > output. All external forces are balanced by the
> > > gastrocnemius
> > > > > > muscles.
> > > > > > > I have checked by varying the force magnitude in
> > > environment.any.
> > > > > > > Using “F=kforce(t)" with varied “k”, I am able to
> change
> > > the
> > > > > > ground
> > > > > > > reaction force input. The output of the forces of
> > > gastrocnemius are
> > > > > > > almost proportional to those in force plate.
> > > > > > >
> > > > > > > 5. In “JointsAndDrivers.any”, I have tested to switch on
> > > > > > Reaction.Type
> > > > > > > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On},
> the
> > > soleus
> > > > > > > forces then appear and gastrocnemius force decrease by
> 1/3.
> > > It
> > > > > > seems
> > > > > > > the problem is at ankle joints.
> > > > > > >
> > > > > > > I appreciate it you can take time to look at my revised
> > > model, and
> > > > > > > give me some further suggestions.
> > > > > > >
> > > > > > > The revised model “Gait3D-rev2” has been up-loaded.
> > > > > > >
> > > > > > > John
> > > > > > >
> > > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > > <support@>
> > > > > > wrote:
> > > > > > > >
> > > > > > > > Hi John
> > > > > > > >
> > > > > > > > I have tested your model and found a few things
> > > > > > > >
> > > > > > > > 1: the position of the CentrOfPressure segment seem to
> be
> > > too
> > > > > > low.
> > > > > > > > This can be corrected by adding a new node in the
> > > globalref
> > > > > > which is
> > > > > > > > located higher, and then refering to this node in the
> > > measures
> > > > > > used
> > > > > > > > for driving this segment.
> > > > > > > >
> > > > > > > > 2: The ankle eversion/inversion seems a bit strange,
> this
> > > might
> > > > > > > > cause problems for the muscles carrying this dof. In
> order
> > > to
> > > > > > see if
> > > > > > > > this is a problem, i inserted these lines in the
> > > > > > > > JointsAndDrivers.any file
> > > > > > > >
> > > > > > > > AnyReacForce AnkleEversion ={ //stc added these
> > > reactions
> > > > > > because
> > > > > > > > AnyKinMeasureOrg
> > > > > > > >
> &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > > AnyKinMeasureOrg
> > > > > > > >
> &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > > };
> > > > > > > >
> > > > > > > >
> > > > > > > > The lines adds a reaction force to ankle
> > > inversion/eversion
> > > > > > dofs. It
> > > > > > > > help the activity to drop a bit, but not enough.
> > > > > > > >
> > > > > > > >
> > > > > > > > 3: Then i switched off the muscles and looked at the
> > > moments.
> > > > > > The
> > > > > > > > left knee moment looks a bit strange you can see by
> > > looking at
> > > > > > the
> > > > > > > > property
> > > > > > > >
> > > > > > >
> > > > > >
> > >
> > "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > > > > of
> > > > > > > > 0.Muscle..Ft”
> > > > > > > >
> > > > > > > > it will give you two curves one of them is the
> negative
> > > moment
> > > > > > the
> > > > > > > > other one the positive moment. Normally the negative
> one
> > > should
> > > > > > be
> > > > > > > > dominant in these models and it should not change sign
> > > during
> > > > > > the
> > > > > > > > stance phase as it does in this case, so i think this
> is
> > > worth
> > > > > > > > investigating a bit. I have not been able to find the
> > > cause of
> > > > > > it,
> > > > > > > > typical reasons could be kinematics of the left leg,
> the
> > > way the
> > > > > > > > force is applied etc.
> > > > > > > >
> > > > > > > > 4: The last thing i have noticed is about the scaling,
> in
> > > the
> > > > > > > > AnyMan.any file it seems that your subject have a
> weigth
> > > of 78
> > > > > > kg,
> > > > > > > > but looking at the segment masses of the thigh and
> shank
> > > they
> > > > > > are
> > > > > > > > unchanged compared to the original Gait3D model. I
> think
> > > these
> > > > > > > > values should be slight higher.
> > > > > > > >
> > > > > > > > I hope it helps you move on, although it do not solve
> the
> > > entire
> > > > > > > > problem.
> > > > > > > >
> > > > > > > > Best regards
> > > > > > > > SÃ¸ren, AnyBody support
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> > > wrote:
> > > > > > > > >
> > > > > > > > > Hi SÃ¸ren:
> > > > > > > > >
> > > > > > > > > I am applying Gait3D using my own data. The model
> seems
> > > > > > working and
> > > > > > > > > the magnitude of the joint moments are also OK. But
> the
> > > > > > predicted
> > > > > > > > > magnitude of muscle forces are very strange.
> Especially,
> > > the
> > > > > > force
> > > > > > > > in
> > > > > > > > > gastrocnemius reaches over 4500 N, it’s too large;
> wile
> > > force
> > > > > > in
> > > > > > > > > soleus is only 200-300N, too small. I have checked
> your
> > > sample
> > > > > > > > Gait3D,
> > > > > > > > > the muscle forces are OK. I have run several other
> data
> > > sets,
> > > > > > and
> > > > > > > > the
> > > > > > > > > results are similar. It seems there is a systematic
> > > problem in
> > > > > > my
> > > > > > > > > model, but I cannot find it. Can you look at my
> model
> > > and give
> > > > > > me
> > > > > > > > some
> > > > > > > > > idea where the problem may be?
> > > > > > > > >
> > > > > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > > > > >
> > > > > > > > > Thanks,
> > > > > > > > >
> > > > > > > > > John Wu
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

system · June 14, 2007, 4:01pm

Hi Soeren:

After a revising of the JointAndDrivers.any, my model works OK. The
kinematics look well.

The muscle force in gastrocnemius of the left leg is still too large.
However, the muscle forces in the right leg seem OK. All my four sets
of data result in similar muscle force outputs. Can you test my model
using your new kinemtics optimizer, to see if it’s possible to improve
the results. If my model is proven to be correct, I will just use the
force output of the right legs. Since the left leg starts first in our
tests, it is resonable to believe that that the movement is not yet
“stablized” at start.

I have up-loaded my model as “Gait3D_rev4.zip”.

Regards,

John

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi Soeren:
>
> Finally I got my model working. It’s now a “clean” solution, without
> using AnkleEversion driver, just driving via motion markers. It seems
> the JointAndDrivers.any in Gait3D does not work on my data. I have
> revised the driver at three driving ponts, and it works well now. The
> gait seems normal. I am still working on it to test my another data
> sets. I have four sets of gait data collected from four different
> subjects. I will let you know the results and what I did.
>
> Thanks for your help,
>
> John
>
> — In anyscript@yahoogroups.com, “AnyBody Support” <support@> wrote:
> >
> > Hi John
> >
> > The kinematic optimizer is working on a model which has a fixed
> > topology, so i can not specify a specific dof like the eversion of
> > the ankle to be rigid unfortunately.
> >
> > You are writing that you have the joint angles measured in Peak
> > system, maybee it would be possible for you to drive the model using
> > these angles directly?. This requires that you know exactly how
> > these angles are measured and then create new rotational measures in
> > AnyBody reflecting this. If these two models match kinematically
> > joint angle drivers should be a good solution.
> >
> > Best regards
> > SÃ¸ren
> >
> > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > >
> > > Hi SÃ¸ren:
> > >
> > > I believe the best way to solve the problem is the kinematics
> > > optimization. The model should be stable, if you can solve the
> > > over-constrained kinemtics using your new optimizer.
> > >
> > > If the problem is related to the ankle eversion/inversion, I have a
> > > better solution. I can drive the ankle eversion directly, while
> > still
> > > driving the model using all other motion markers. We have data for
> > all
> > > joint angles measured using Peak system. Could you try it using
> > your
> > > kinematics optimizer? If it does not work, please try to put ankle
> > > eversion to zero. I have up-loaded the ankle-eversion data and the
> > > revised JointAndDrivers.any. (Gait3D_ankle_eversion.zip)
> > >
> > > The unsymmetric model seems too complicated. And I am not sure if
> > my
> > > model will work, putting so much effects to change the human model
> > to
> > > unsymmetric. So, I would not proceed in that direction.
> > >
> > > Thanks for your effects,
> > >
> > > John
> > >
> > > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> > wrote:
> > > >
> > > > Hi John
> > > >
> > > > I have tried to optimize the kinematics using an optimizer made
> > by
> > > > Michael Skipper Andersen, who is working on a Ph.D. project,
> > about
> > > > kinematic analysis of over-determinate systems. Unfortunately I
> > have
> > > > not been doing successful doing so yet. The reason is the feet;
> > > > since there are only two markers here it creates problems for
> > the
> > > > algorithm to determine primarily the eversion/inversion dof, in
> > a
> > > > good way. Especially the left foot is problematic because the
> > heel,
> > > > mallous and metatarsal markers are almost lying on straight
> > line.
> > > > This problem will persist regardless of which method we use to
> > > > impose the movement on the model. If you a have a dataset with
> > more
> > > > markers on the forefoot it would be easier for the algorithm to
> > > > solve this problem.
> > > >
> > > > You are right that improving the kinematics manually is rather
> > > > difficult and time consuming; this was why I tried using
> > Michael’s
> > > > optimizer on your problem. If it had been successful the output
> > > > would have been the optimized marker positions, segments
> > lengths,
> > > > and joint angles, but I am afraid that for the current data set
> > this
> > > > is not possible, I had to abandon this strategy after having
> > spend
> > > > quite some time to make it work.
> > > >
> > > > It is possible that scaling the left and right side differently
> > will
> > > > improve this model. This can not be done straight away and
> > requires
> > > > some changes to be made in the BRep.
> > > >
> > > > I will try list the needed changes here:
> > > > 1. In the AnyMan you need to make a left and right side folder
> > > > obviously, containing the segment lengths for the two sides.
> > > > 2. In the UniformScaling file you need to create a left and
> > right
> > > > side scaling for the legs segments.
> > > > 3. In the BodyModels directory in the specific BodyModel file
> > you
> > > > are using you need to introduce a new folder reference, for
> > example
> > > > named ScalingRightLeft, in the Right and Left side folders. This
> > > > reference should point at the scaling laws just created. This
> > folder
> > > > reference should have the same name in both folders, but point
> > at
> > > > different scalinglaws.
> > > > 4. In the Leg3d/Seg…any file you need to point at the scaling
> > law
> > > > defined in the bodymodel right and left folder., instead of the
> > > > existing scaling which are there today.
> > > >
> > > > This should introduce different scaling to the right and left
> > side.
> > > >
> > > > I hope scaling the left and right side differently will help
> > solve
> > > > the problem, but the inversion and eversion of the foot will
> > > > continue to be a problem.
> > > >
> > > > Best regards
> > > > SÃ¸ren, AnyBody Support
> > > >
> > > >
> > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > >
> > > > > Hi SÃ¸ren:
> > > > >
> > > > > I have improved the model a little bit. It can now run to the
> > end.
> > > > > But, the kinematics is still a problem. The updated model is
> > > > > Gait3D_rev3, which is up-loaded.
> > > > >
> > > > > John
> > > > >
> > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > > >
> > > > > > Hi SÃ¸ren:
> > > > > >
> > > > > > Thanks for your time and efforts. I agree with you that the
> > > > problem is
> > > > > > likely caused by the kinematics. But, it is a tough job. I
> > have
> > > > spent
> > > > > > a lot of time to have my model scaled and motion markers
> > > > adjusted.
> > > > > > When it is initialized, the model looks very good. But, it
> > runs
> > > > only
> > > > > > 20 steps and stops. What I found is that the subject in our
> > > > study is
> > > > > > unsymmetric. It is shown in my model, if you look at it
> > closely.
> > > > The
> > > > > > thigh and shank lengths are different from left to right. We
> > also
> > > > > > found it in our direct physical measurements. This may cause
> > the
> > > > > > problem. My questions is: how to adjust the left and right
> > thigh
> > > > and
> > > > > > shank length separately?
> > > > > >
> > > > > > I have revised the MotionMarkerDrivers, but it does not help.
> > > > > >
> > > > > > I have up-loaded my up-dated model “Gait3D-rev3”. All motion
> > > > markers,
> > > > > > which are not used to drive the model, have been removed in
> > the
> > > > model
> > > > > > to make it clearere to view.
> > > > > >
> > > > > > Thanks,
> > > > > >
> > > > > > John
> > > > > >
> > > > > >
> > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > <support@>
> > > > wrote:
> > > > > > >
> > > > > > > Hi John
> > > > > > >
> > > > > > > I have looked a bit more into the model
> > > > > > >
> > > > > > >
> > > > > > > I think the main problem could be the kinemtatics. I
> > tried
> > > > again to
> > > > > > > insert a reaction force on the right and left ankle
> > > > > > > inversion/eversion and the activity dropped from 15 to
> > 5.8.
> > > > This
> > > > > > > might be cause by the angle between the foot and the floor
> > > > (tilt), it
> > > > > > > looks like he is walking on the lateral side of the feets.
> > > > > > >
> > > > > > > Secondly i looked at the motion of the pelvis, if you look
> > the
> > > > > > > motion from the front there are some very large rotations
> > of
> > > > the
> > > > > > > pelvis, and it can be seen that there is a large error on
> > the
> > > > Asis
> > > > > > > markers. Therefore i tried adding reaction forces to the
> > hip
> > > > > > > abduction and external rotatation together with the ankle
> > > > inversion
> > > > > > > moment.
> > > > > > >
> > > > > > >
> > > > > > >
> > > > > > > AnyReacForce LeftSideJoints ={
> > > > > > >
> > > > > > > AnyKinMeasureOrg &ref2=
> > > > > > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > AnyKinMeasureOrg &ref5=
> > > > > > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > > > > > AnyKinMeasureOrg &ref6=
> > > > > > > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > > > > > >
> > > > > > > };
> > > > > > >
> > > > > > >
> > > > > > > AnyReacForce RightSideJoints ={
> > > > > > >
> > > > > > > AnyKinMeasureOrg &ref2=
> > > > > > > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > AnyKinMeasureOrg &ref5=
> > > > > > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > > > > > AnyKinMeasureOrg &ref6=
> > > > > > > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > > > > > >
> > > > > > > };
> > > > > > >
> > > > > > > This helped the activity to drop to 1.4. I think the
> > > > explanation is
> > > > > > > that the momentarms around the hip joint becomes too small
> > for
> > > > this
> > > > > > > amount of pelvis “tilt”.
> > > > > > >
> > > > > > > So i think the next step will be to adjust the motion.
> > > > > > >
> > > > > > >
> > > > > > > I looked at the moment curves and compared them to the
> > curves
> > > > which
> > > > > > > can be found in book "Dynamics of human gait "
> > > > > > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > > > > > >
> > > > > > > On page 42 there is a curve for the moment around the hip,
> > > > knee and
> > > > > > > ankle, when i compare the knee moment with the
> > > > > > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks quite
> > > > different?
> > > > > > >
> > > > > > > My impression is that it is the kinematics which are
> > causing
> > > > these
> > > > > > > problems primarily. I think the scaling and the marker
> > > > postions
> > > > > > > should be adjusted until the motion gives a smaller
> > hip “tilt”
> > > > and
> > > > > > > the foot tilt are more horizontal and aligned with the
> > floor.
> > > > > > >
> > > > > > >
> > > > > > > Best regards
> > > > > > > SÃ¸ren, AnyBody support
> > > > > > >
> > > > > > >
> > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> > wrote:
> > > > > > > >
> > > > > > > > Hi SÃ¸ren:
> > > > > > > >
> > > > > > > > Thanks for your suggestions. I have improved my model,
> > as
> > > > > > > suggested.
> > > > > > > >
> > > > > > > > 1. The CenterOfPressure drivers have been revised. Now,
> > I
> > > > drive
> > > > > > > > CenterOfPressure only in x-y plane, and raise it to a
> > height
> > > > > > > of “pz”,
> > > > > > > > which can ben adjusted. By doing this, the muscle force
> > in
> > > > > > > > gastrocnemius is reduced by 10%, still too high. (See my
> > > > revised
> > > > > > > > “enviroment.any” file.)
> > > > > > > >
> > > > > > > > 2. All section masses have been scaled. But this has
> > little
> > > > > > > influence.
> > > > > > > > (see revised “man.any”)
> > > > > > > >
> > > > > > > > 3. I have checked the joint moment by using “No_Muscle”
> > and
> > > > output
> > > > > > > the
> > > > > > > > joint muscle force. The sum of plus and minus force
> > gives
> > > > the
> > > > > > > moment.
> > > > > > > > I have added a data processing unit
> > (“Output_kinetic.any”)
> > > > to
> > > > > > > output
> > > > > > > > the joint moment. Note: M and nM respresent moment and
> > > > normalized
> > > > > > > > moment, respectively. When normalized to the body mass,
> > the
> > > > joint
> > > > > > > > moments are consistent with published data.
> > > > > > > >
> > > > > > > > 4. I found that the soleus muscle seems not active and
> > has
> > > > no force
> > > > > > > > output. All external forces are balanced by the
> > > > gastrocnemius
> > > > > > > muscles.
> > > > > > > > I have checked by varying the force magnitude in
> > > > environment.any.
> > > > > > > > Using “F=kforce(t)" with varied “k”, I am able to
> > change
> > > > the
> > > > > > > ground
> > > > > > > > reaction force input. The output of the forces of
> > > > gastrocnemius are
> > > > > > > > almost proportional to those in force plate.
> > > > > > > >
> > > > > > > > 5. In “JointsAndDrivers.any”, I have tested to switch on
> > > > > > > Reaction.Type
> > > > > > > > at HeelYLDrv and HeelYLDrv to Reaction.Type={Off,On,On},
> > the
> > > > soleus
> > > > > > > > forces then appear and gastrocnemius force decrease by
> > 1/3.
> > > > It
> > > > > > > seems
> > > > > > > > the problem is at ankle joints.
> > > > > > > >
> > > > > > > > I appreciate it you can take time to look at my revised
> > > > model, and
> > > > > > > > give me some further suggestions.
> > > > > > > >
> > > > > > > > The revised model “Gait3D-rev2” has been up-loaded.
> > > > > > > >
> > > > > > > > John
> > > > > > > >
> > > > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > > > <support@>
> > > > > > > wrote:
> > > > > > > > >
> > > > > > > > > Hi John
> > > > > > > > >
> > > > > > > > > I have tested your model and found a few things
> > > > > > > > >
> > > > > > > > > 1: the position of the CentrOfPressure segment seem to
> > be
> > > > too
> > > > > > > low.
> > > > > > > > > This can be corrected by adding a new node in the
> > > > globalref
> > > > > > > which is
> > > > > > > > > located higher, and then refering to this node in the
> > > > measures
> > > > > > > used
> > > > > > > > > for driving this segment.
> > > > > > > > >
> > > > > > > > > 2: The ankle eversion/inversion seems a bit strange,
> > this
> > > > might
> > > > > > > > > cause problems for the muscles carrying this dof. In
> > order
> > > > to
> > > > > > > see if
> > > > > > > > > this is a problem, i inserted these lines in the
> > > > > > > > > JointsAndDrivers.any file
> > > > > > > > >
> > > > > > > > > AnyReacForce AnkleEversion ={ //stc added these
> > > > reactions
> > > > > > > because
> > > > > > > > > AnyKinMeasureOrg
> > > > > > > > >
> > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > > > AnyKinMeasureOrg
> > > > > > > > >
> > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > > > };
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > The lines adds a reaction force to ankle
> > > > inversion/eversion
> > > > > > > dofs. It
> > > > > > > > > help the activity to drop a bit, but not enough.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > 3: Then i switched off the muscles and looked at the
> > > > moments.
> > > > > > > The
> > > > > > > > > left knee moment looks a bit strange you can see by
> > > > looking at
> > > > > > > the
> > > > > > > > > property
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > >
> > > "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > > > > > of
> > > > > > > > > 0.Muscle..Ft”
> > > > > > > > >
> > > > > > > > > it will give you two curves one of them is the
> > negative
> > > > moment
> > > > > > > the
> > > > > > > > > other one the positive moment. Normally the negative
> > one
> > > > should
> > > > > > > be
> > > > > > > > > dominant in these models and it should not change sign
> > > > during
> > > > > > > the
> > > > > > > > > stance phase as it does in this case, so i think this
> > is
> > > > worth
> > > > > > > > > investigating a bit. I have not been able to find the
> > > > cause of
> > > > > > > it,
> > > > > > > > > typical reasons could be kinematics of the left leg,
> > the
> > > > way the
> > > > > > > > > force is applied etc.
> > > > > > > > >
> > > > > > > > > 4: The last thing i have noticed is about the scaling,
> > in
> > > > the
> > > > > > > > > AnyMan.any file it seems that your subject have a
> > weigth
> > > > of 78
> > > > > > > kg,
> > > > > > > > > but looking at the segment masses of the thigh and
> > shank
> > > > they
> > > > > > > are
> > > > > > > > > unchanged compared to the original Gait3D model. I
> > think
> > > > these
> > > > > > > > > values should be slight higher.
> > > > > > > > >
> > > > > > > > > I hope it helps you move on, although it do not solve
> > the
> > > > entire
> > > > > > > > > problem.
> > > > > > > > >
> > > > > > > > > Best regards
> > > > > > > > > SÃ¸ren, AnyBody support
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> > > > wrote:
> > > > > > > > > >
> > > > > > > > > > Hi SÃ¸ren:
> > > > > > > > > >
> > > > > > > > > > I am applying Gait3D using my own data. The model
> > seems
> > > > > > > working and
> > > > > > > > > > the magnitude of the joint moments are also OK. But
> > the
> > > > > > > predicted
> > > > > > > > > > magnitude of muscle forces are very strange.
> > Especially,
> > > > the
> > > > > > > force
> > > > > > > > > in
> > > > > > > > > > gastrocnemius reaches over 4500 N, it’s too large;
> > wile
> > > > force
> > > > > > > in
> > > > > > > > > > soleus is only 200-300N, too small. I have checked
> > your
> > > > sample
> > > > > > > > > Gait3D,
> > > > > > > > > > the muscle forces are OK. I have run several other
> > data
> > > > sets,
> > > > > > > and
> > > > > > > > > the
> > > > > > > > > > results are similar. It seems there is a systematic
> > > > problem in
> > > > > > > my
> > > > > > > > > > model, but I cannot find it. Can you look at my
> > model
> > > > and give
> > > > > > > me
> > > > > > > > > some
> > > > > > > > > > idea where the problem may be?
> > > > > > > > > >
> > > > > > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > > > > > >
> > > > > > > > > > Thanks,
> > > > > > > > > >
> > > > > > > > > > John Wu
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

system · June 17, 2007, 5:01am

Hi Soeren:

I have found out that there are some problems with my motion data. The
horizontal force is too large compared to the vertical force.
Comparing our data with your sample Gait3D, I found: the ratio of the
peak force magnitude of the horzontal (x-axis) to vertical (z-axis) in
Gait3D sample is 1:6, while it is about 1:2 in our gait motion data.
By scaling of the horizontal force to a factor of 0.33, I obtained the
forces in soleus and gastrocnemius comparable to those obtained in
Gait3D-sample. My gait model seems fine. I will check the calibration
of the force plate in our lab next week.

I really appreciate your great suggestions and efforts.

John Wu

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi Soeren:
>
> After a revising of the JointAndDrivers.any, my model works OK. The
> kinematics look well.
>
> The muscle force in gastrocnemius of the left leg is still too large.
> However, the muscle forces in the right leg seem OK. All my four sets
> of data result in similar muscle force outputs. Can you test my model
> using your new kinemtics optimizer, to see if it’s possible to improve
> the results. If my model is proven to be correct, I will just use the
> force output of the right legs. Since the left leg starts first in our
> tests, it is resonable to believe that that the movement is not yet
> “stablized” at start.
>
> I have up-loaded my model as “Gait3D_rev4.zip”.
>
> Regards,
>
> John
>
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi Soeren:
> >
> > Finally I got my model working. It’s now a “clean” solution, without
> > using AnkleEversion driver, just driving via motion markers. It seems
> > the JointAndDrivers.any in Gait3D does not work on my data. I have
> > revised the driver at three driving ponts, and it works well now. The
> > gait seems normal. I am still working on it to test my another data
> > sets. I have four sets of gait data collected from four different
> > subjects. I will let you know the results and what I did.
> >
> > Thanks for your help,
> >
> > John
> >
> > — In anyscript@yahoogroups.com, “AnyBody Support” <support@> wrote:
> > >
> > > Hi John
> > >
> > > The kinematic optimizer is working on a model which has a fixed
> > > topology, so i can not specify a specific dof like the eversion of
> > > the ankle to be rigid unfortunately.
> > >
> > > You are writing that you have the joint angles measured in Peak
> > > system, maybee it would be possible for you to drive the model
using
> > > these angles directly?. This requires that you know exactly how
> > > these angles are measured and then create new rotational
measures in
> > > AnyBody reflecting this. If these two models match kinematically
> > > joint angle drivers should be a good solution.
> > >
> > > Best regards
> > > SÃ¸ren
> > >
> > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > >
> > > > Hi SÃ¸ren:
> > > >
> > > > I believe the best way to solve the problem is the kinematics
> > > > optimization. The model should be stable, if you can solve the
> > > > over-constrained kinemtics using your new optimizer.
> > > >
> > > > If the problem is related to the ankle eversion/inversion, I
have a
> > > > better solution. I can drive the ankle eversion directly, while
> > > still
> > > > driving the model using all other motion markers. We have data
for
> > > all
> > > > joint angles measured using Peak system. Could you try it using
> > > your
> > > > kinematics optimizer? If it does not work, please try to put ankle
> > > > eversion to zero. I have up-loaded the ankle-eversion data and the
> > > > revised JointAndDrivers.any. (Gait3D_ankle_eversion.zip)
> > > >
> > > > The unsymmetric model seems too complicated. And I am not sure if
> > > my
> > > > model will work, putting so much effects to change the human
model
> > > to
> > > > unsymmetric. So, I would not proceed in that direction.
> > > >
> > > > Thanks for your effects,
> > > >
> > > > John
> > > >
> > > > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
> > > wrote:
> > > > >
> > > > > Hi John
> > > > >
> > > > > I have tried to optimize the kinematics using an optimizer made
> > > by
> > > > > Michael Skipper Andersen, who is working on a Ph.D. project,
> > > about
> > > > > kinematic analysis of over-determinate systems. Unfortunately I
> > > have
> > > > > not been doing successful doing so yet. The reason is the feet;
> > > > > since there are only two markers here it creates problems for
> > > the
> > > > > algorithm to determine primarily the eversion/inversion dof, in
> > > a
> > > > > good way. Especially the left foot is problematic because the
> > > heel,
> > > > > mallous and metatarsal markers are almost lying on straight
> > > line.
> > > > > This problem will persist regardless of which method we use to
> > > > > impose the movement on the model. If you a have a dataset with
> > > more
> > > > > markers on the forefoot it would be easier for the algorithm to
> > > > > solve this problem.
> > > > >
> > > > > You are right that improving the kinematics manually is rather
> > > > > difficult and time consuming; this was why I tried using
> > > Michael’s
> > > > > optimizer on your problem. If it had been successful the output
> > > > > would have been the optimized marker positions, segments
> > > lengths,
> > > > > and joint angles, but I am afraid that for the current data set
> > > this
> > > > > is not possible, I had to abandon this strategy after having
> > > spend
> > > > > quite some time to make it work.
> > > > >
> > > > > It is possible that scaling the left and right side differently
> > > will
> > > > > improve this model. This can not be done straight away and
> > > requires
> > > > > some changes to be made in the BRep.
> > > > >
> > > > > I will try list the needed changes here:
> > > > > 1. In the AnyMan you need to make a left and right side folder
> > > > > obviously, containing the segment lengths for the two sides.
> > > > > 2. In the UniformScaling file you need to create a left and
> > > right
> > > > > side scaling for the legs segments.
> > > > > 3. In the BodyModels directory in the specific BodyModel file
> > > you
> > > > > are using you need to introduce a new folder reference, for
> > > example
> > > > > named ScalingRightLeft, in the Right and Left side folders.
This
> > > > > reference should point at the scaling laws just created. This
> > > folder
> > > > > reference should have the same name in both folders, but point
> > > at
> > > > > different scalinglaws.
> > > > > 4. In the Leg3d/Seg…any file you need to point at the scaling
> > > law
> > > > > defined in the bodymodel right and left folder., instead of the
> > > > > existing scaling which are there today.
> > > > >
> > > > > This should introduce different scaling to the right and left
> > > side.
> > > > >
> > > > > I hope scaling the left and right side differently will help
> > > solve
> > > > > the problem, but the inversion and eversion of the foot will
> > > > > continue to be a problem.
> > > > >
> > > > > Best regards
> > > > > SÃ¸ren, AnyBody Support
> > > > >
> > > > >
> > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > > >
> > > > > > Hi SÃ¸ren:
> > > > > >
> > > > > > I have improved the model a little bit. It can now run to the
> > > end.
> > > > > > But, the kinematics is still a problem. The updated model is
> > > > > > Gait3D_rev3, which is up-loaded.
> > > > > >
> > > > > > John
> > > > > >
> > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > > > >
> > > > > > > Hi SÃ¸ren:
> > > > > > >
> > > > > > > Thanks for your time and efforts. I agree with you that the
> > > > > problem is
> > > > > > > likely caused by the kinematics. But, it is a tough job. I
> > > have
> > > > > spent
> > > > > > > a lot of time to have my model scaled and motion markers
> > > > > adjusted.
> > > > > > > When it is initialized, the model looks very good. But, it
> > > runs
> > > > > only
> > > > > > > 20 steps and stops. What I found is that the subject in our
> > > > > study is
> > > > > > > unsymmetric. It is shown in my model, if you look at it
> > > closely.
> > > > > The
> > > > > > > thigh and shank lengths are different from left to
right. We
> > > also
> > > > > > > found it in our direct physical measurements. This may
cause
> > > the
> > > > > > > problem. My questions is: how to adjust the left and right
> > > thigh
> > > > > and
> > > > > > > shank length separately?
> > > > > > >
> > > > > > > I have revised the MotionMarkerDrivers, but it does not
help.
> > > > > > >
> > > > > > > I have up-loaded my up-dated model “Gait3D-rev3”. All
motion
> > > > > markers,
> > > > > > > which are not used to drive the model, have been removed in
> > > the
> > > > > model
> > > > > > > to make it clearere to view.
> > > > > > >
> > > > > > > Thanks,
> > > > > > >
> > > > > > > John
> > > > > > >
> > > > > > >
> > > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > > <support@>
> > > > > wrote:
> > > > > > > >
> > > > > > > > Hi John
> > > > > > > >
> > > > > > > > I have looked a bit more into the model
> > > > > > > >
> > > > > > > >
> > > > > > > > I think the main problem could be the kinemtatics. I
> > > tried
> > > > > again to
> > > > > > > > insert a reaction force on the right and left ankle
> > > > > > > > inversion/eversion and the activity dropped from 15 to
> > > 5.8.
> > > > > This
> > > > > > > > might be cause by the angle between the foot and the floor
> > > > > (tilt), it
> > > > > > > > looks like he is walking on the lateral side of the
feets.
> > > > > > > >
> > > > > > > > Secondly i looked at the motion of the pelvis, if you
look
> > > the
> > > > > > > > motion from the front there are some very large rotations
> > > of
> > > > > the
> > > > > > > > pelvis, and it can be seen that there is a large error on
> > > the
> > > > > Asis
> > > > > > > > markers. Therefore i tried adding reaction forces to the
> > > hip
> > > > > > > > abduction and external rotatation together with the ankle
> > > > > inversion
> > > > > > > > moment.
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > AnyReacForce LeftSideJoints ={
> > > > > > > >
> > > > > > > > AnyKinMeasureOrg &ref2=
> > > > > > > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > > AnyKinMeasureOrg &ref5=
> > > > > > > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > > > > > > AnyKinMeasureOrg &ref6=
> > > > > > > > Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > > > > > > >
> > > > > > > > };
> > > > > > > >
> > > > > > > >
> > > > > > > > AnyReacForce RightSideJoints ={
> > > > > > > >
> > > > > > > > AnyKinMeasureOrg &ref2=
> > > > > > > > Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > > AnyKinMeasureOrg &ref5=
> > > > > > > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > > > > > > AnyKinMeasureOrg &ref6=
> > > > > > > > Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > > > > > > >
> > > > > > > > };
> > > > > > > >
> > > > > > > > This helped the activity to drop to 1.4. I think the
> > > > > explanation is
> > > > > > > > that the momentarms around the hip joint becomes too
small
> > > for
> > > > > this
> > > > > > > > amount of pelvis “tilt”.
> > > > > > > >
> > > > > > > > So i think the next step will be to adjust the motion.
> > > > > > > >
> > > > > > > >
> > > > > > > > I looked at the moment curves and compared them to the
> > > curves
> > > > > which
> > > > > > > > can be found in book "Dynamics of human gait "
> > > > > > > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > > > > > > >
> > > > > > > > On page 42 there is a curve for the moment around the
hip,
> > > > > knee and
> > > > > > > > ankle, when i compare the knee moment with the
> > > > > > > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it looks
quite
> > > > > different?
> > > > > > > >
> > > > > > > > My impression is that it is the kinematics which are
> > > causing
> > > > > these
> > > > > > > > problems primarily. I think the scaling and the marker
> > > > > postions
> > > > > > > > should be adjusted until the motion gives a smaller
> > > hip “tilt”
> > > > > and
> > > > > > > > the foot tilt are more horizontal and aligned with the
> > > floor.
> > > > > > > >
> > > > > > > >
> > > > > > > > Best regards
> > > > > > > > SÃ¸ren, AnyBody support
> > > > > > > >
> > > > > > > >
> > > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> > > wrote:
> > > > > > > > >
> > > > > > > > > Hi SÃ¸ren:
> > > > > > > > >
> > > > > > > > > Thanks for your suggestions. I have improved my model,
> > > as
> > > > > > > > suggested.
> > > > > > > > >
> > > > > > > > > 1. The CenterOfPressure drivers have been revised. Now,
> > > I
> > > > > drive
> > > > > > > > > CenterOfPressure only in x-y plane, and raise it to a
> > > height
> > > > > > > > of “pz”,
> > > > > > > > > which can ben adjusted. By doing this, the muscle force
> > > in
> > > > > > > > > gastrocnemius is reduced by 10%, still too high.
(See my
> > > > > revised
> > > > > > > > > “enviroment.any” file.)
> > > > > > > > >
> > > > > > > > > 2. All section masses have been scaled. But this has
> > > little
> > > > > > > > influence.
> > > > > > > > > (see revised “man.any”)
> > > > > > > > >
> > > > > > > > > 3. I have checked the joint moment by using “No_Muscle”
> > > and
> > > > > output
> > > > > > > > the
> > > > > > > > > joint muscle force. The sum of plus and minus force
> > > gives
> > > > > the
> > > > > > > > moment.
> > > > > > > > > I have added a data processing unit
> > > (“Output_kinetic.any”)
> > > > > to
> > > > > > > > output
> > > > > > > > > the joint moment. Note: M and nM respresent moment and
> > > > > normalized
> > > > > > > > > moment, respectively. When normalized to the body mass,
> > > the
> > > > > joint
> > > > > > > > > moments are consistent with published data.
> > > > > > > > >
> > > > > > > > > 4. I found that the soleus muscle seems not active and
> > > has
> > > > > no force
> > > > > > > > > output. All external forces are balanced by the
> > > > > gastrocnemius
> > > > > > > > muscles.
> > > > > > > > > I have checked by varying the force magnitude in
> > > > > environment.any.
> > > > > > > > > Using “F=kforce(t)" with varied “k”, I am able to
> > > change
> > > > > the
> > > > > > > > ground
> > > > > > > > > reaction force input. The output of the forces of
> > > > > gastrocnemius are
> > > > > > > > > almost proportional to those in force plate.
> > > > > > > > >
> > > > > > > > > 5. In “JointsAndDrivers.any”, I have tested to
switch on
> > > > > > > > Reaction.Type
> > > > > > > > > at HeelYLDrv and HeelYLDrv to
Reaction.Type={Off,On,On},
> > > the
> > > > > soleus
> > > > > > > > > forces then appear and gastrocnemius force decrease by
> > > 1/3.
> > > > > It
> > > > > > > > seems
> > > > > > > > > the problem is at ankle joints.
> > > > > > > > >
> > > > > > > > > I appreciate it you can take time to look at my revised
> > > > > model, and
> > > > > > > > > give me some further suggestions.
> > > > > > > > >
> > > > > > > > > The revised model “Gait3D-rev2” has been up-loaded.
> > > > > > > > >
> > > > > > > > > John
> > > > > > > > >
> > > > > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > > > > <support@>
> > > > > > > > wrote:
> > > > > > > > > >
> > > > > > > > > > Hi John
> > > > > > > > > >
> > > > > > > > > > I have tested your model and found a few things
> > > > > > > > > >
> > > > > > > > > > 1: the position of the CentrOfPressure segment
seem to
> > > be
> > > > > too
> > > > > > > > low.
> > > > > > > > > > This can be corrected by adding a new node in the
> > > > > globalref
> > > > > > > > which is
> > > > > > > > > > located higher, and then refering to this node in the
> > > > > measures
> > > > > > > > used
> > > > > > > > > > for driving this segment.
> > > > > > > > > >
> > > > > > > > > > 2: The ankle eversion/inversion seems a bit strange,
> > > this
> > > > > might
> > > > > > > > > > cause problems for the muscles carrying this dof. In
> > > order
> > > > > to
> > > > > > > > see if
> > > > > > > > > > this is a problem, i inserted these lines in the
> > > > > > > > > > JointsAndDrivers.any file
> > > > > > > > > >
> > > > > > > > > > AnyReacForce AnkleEversion ={ //stc added these
> > > > > reactions
> > > > > > > > because
> > > > > > > > > > AnyKinMeasureOrg
> > > > > > > > > >
> > > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > > > > AnyKinMeasureOrg
> > > > > > > > > >
> > > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > > > > };
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > The lines adds a reaction force to ankle
> > > > > inversion/eversion
> > > > > > > > dofs. It
> > > > > > > > > > help the activity to drop a bit, but not enough.
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > 3: Then i switched off the muscles and looked at the
> > > > > moments.
> > > > > > > > The
> > > > > > > > > > left knee moment looks a bit strange you can see by
> > > > > looking at
> > > > > > > > the
> > > > > > > > > > property
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > >
> > > >
"Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > > > > > > of
> > > > > > > > > > 0.Muscle..Ft”
> > > > > > > > > >
> > > > > > > > > > it will give you two curves one of them is the
> > > negative
> > > > > moment
> > > > > > > > the
> > > > > > > > > > other one the positive moment. Normally the negative
> > > one
> > > > > should
> > > > > > > > be
> > > > > > > > > > dominant in these models and it should not change
sign
> > > > > during
> > > > > > > > the
> > > > > > > > > > stance phase as it does in this case, so i think this
> > > is
> > > > > worth
> > > > > > > > > > investigating a bit. I have not been able to find the
> > > > > cause of
> > > > > > > > it,
> > > > > > > > > > typical reasons could be kinematics of the left leg,
> > > the
> > > > > way the
> > > > > > > > > > force is applied etc.
> > > > > > > > > >
> > > > > > > > > > 4: The last thing i have noticed is about the
scaling,
> > > in
> > > > > the
> > > > > > > > > > AnyMan.any file it seems that your subject have a
> > > weigth
> > > > > of 78
> > > > > > > > kg,
> > > > > > > > > > but looking at the segment masses of the thigh and
> > > shank
> > > > > they
> > > > > > > > are
> > > > > > > > > > unchanged compared to the original Gait3D model. I
> > > think
> > > > > these
> > > > > > > > > > values should be slight higher.
> > > > > > > > > >
> > > > > > > > > > I hope it helps you move on, although it do not solve
> > > the
> > > > > entire
> > > > > > > > > > problem.
> > > > > > > > > >
> > > > > > > > > > Best regards
> > > > > > > > > > SÃ¸ren, AnyBody support
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > >
> > > > > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
> > > > > wrote:
> > > > > > > > > > >
> > > > > > > > > > > Hi SÃ¸ren:
> > > > > > > > > > >
> > > > > > > > > > > I am applying Gait3D using my own data. The model
> > > seems
> > > > > > > > working and
> > > > > > > > > > > the magnitude of the joint moments are also OK. But
> > > the
> > > > > > > > predicted
> > > > > > > > > > > magnitude of muscle forces are very strange.
> > > Especially,
> > > > > the
> > > > > > > > force
> > > > > > > > > > in
> > > > > > > > > > > gastrocnemius reaches over 4500 N, it’s too large;
> > > wile
> > > > > force
> > > > > > > > in
> > > > > > > > > > > soleus is only 200-300N, too small. I have checked
> > > your
> > > > > sample
> > > > > > > > > > Gait3D,
> > > > > > > > > > > the muscle forces are OK. I have run several other
> > > data
> > > > > sets,
> > > > > > > > and
> > > > > > > > > > the
> > > > > > > > > > > results are similar. It seems there is a systematic
> > > > > problem in
> > > > > > > > my
> > > > > > > > > > > model, but I cannot find it. Can you look at my
> > > model
> > > > > and give
> > > > > > > > me
> > > > > > > > > > some
> > > > > > > > > > > idea where the problem may be?
> > > > > > > > > > >
> > > > > > > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > > > > > > >
> > > > > > > > > > > Thanks,
> > > > > > > > > > >
> > > > > > > > > > > John Wu
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

system · June 17, 2007, 1:09pm

Hi John

It is great to hear that the model is working now, i am sorry that
your last posting was not ansewered, but you seem to have solved the
problems now

Please let me know if you have more questions.

Thanks for the kind words.

Best regards
SÃ¸ren, AnyBody support

— In anyscript@yahoogroups.com, “johnzengwu” <jwu@…> wrote:
>
> Hi Soeren:
>
> I have found out that there are some problems with my motion data.
The
> horizontal force is too large compared to the vertical force.
> Comparing our data with your sample Gait3D, I found: the ratio of
the
> peak force magnitude of the horzontal (x-axis) to vertical (z-
axis) in
> Gait3D sample is 1:6, while it is about 1:2 in our gait motion
data.
> By scaling of the horizontal force to a factor of 0.33, I obtained
the
> forces in soleus and gastrocnemius comparable to those obtained in
> Gait3D-sample. My gait model seems fine. I will check the
calibration
> of the force plate in our lab next week.
>
> I really appreciate your great suggestions and efforts.
>
> John Wu
>
> — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> >
> > Hi Soeren:
> >
> > After a revising of the JointAndDrivers.any, my model works OK.
The
> > kinematics look well.
> >
> > The muscle force in gastrocnemius of the left leg is still too
large.
> > However, the muscle forces in the right leg seem OK. All my four
sets
> > of data result in similar muscle force outputs. Can you test my
model
> > using your new kinemtics optimizer, to see if it’s possible to
improve
> > the results. If my model is proven to be correct, I will just
use the
> > force output of the right legs. Since the left leg starts first
in our
> > tests, it is resonable to believe that that the movement is not
yet
> > “stablized” at start.
> >
> > I have up-loaded my model as “Gait3D_rev4.zip”.
> >
> > Regards,
> >
> > John
> >
> >
> > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > >
> > > Hi Soeren:
> > >
> > > Finally I got my model working. It’s now a “clean” solution,
without
> > > using AnkleEversion driver, just driving via motion markers.
It seems
> > > the JointAndDrivers.any in Gait3D does not work on my data. I
have
> > > revised the driver at three driving ponts, and it works well
now. The
> > > gait seems normal. I am still working on it to test my another
data
> > > sets. I have four sets of gait data collected from four
different
> > > subjects. I will let you know the results and what I did.
> > >
> > > Thanks for your help,
> > >
> > > John
> > >
> > > — In anyscript@yahoogroups.com, “AnyBody Support” <support@>
wrote:
> > > >
> > > > Hi John
> > > >
> > > > The kinematic optimizer is working on a model which has a
fixed
> > > > topology, so i can not specify a specific dof like the
eversion of
> > > > the ankle to be rigid unfortunately.
> > > >
> > > > You are writing that you have the joint angles measured in
Peak
> > > > system, maybee it would be possible for you to drive the
model
> using
> > > > these angles directly?. This requires that you know exactly
how
> > > > these angles are measured and then create new rotational
> measures in
> > > > AnyBody reflecting this. If these two models match
kinematically
> > > > joint angle drivers should be a good solution.
> > > >
> > > > Best regards
> > > > SÃ¸ren
> > > >
> > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@> wrote:
> > > > >
> > > > > Hi SÃ¸ren:
> > > > >
> > > > > I believe the best way to solve the problem is the
kinematics
> > > > > optimization. The model should be stable, if you can solve
the
> > > > > over-constrained kinemtics using your new optimizer.
> > > > >
> > > > > If the problem is related to the ankle eversion/inversion,
I
> have a
> > > > > better solution. I can drive the ankle eversion directly,
while
> > > > still
> > > > > driving the model using all other motion markers. We have
data
> for
> > > > all
> > > > > joint angles measured using Peak system. Could you try it
using
> > > > your
> > > > > kinematics optimizer? If it does not work, please try to
put ankle
> > > > > eversion to zero. I have up-loaded the ankle-eversion data
and the
> > > > > revised JointAndDrivers.any. (Gait3D_ankle_eversion.zip)
> > > > >
> > > > > The unsymmetric model seems too complicated. And I am not
sure if
> > > > my
> > > > > model will work, putting so much effects to change the
human
> model
> > > > to
> > > > > unsymmetric. So, I would not proceed in that direction.
> > > > >
> > > > > Thanks for your effects,
> > > > >
> > > > > John
> > > > >
> > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
<support@>
> > > > wrote:
> > > > > >
> > > > > > Hi John
> > > > > >
> > > > > > I have tried to optimize the kinematics using an
optimizer made
> > > > by
> > > > > > Michael Skipper Andersen, who is working on a Ph.D.
project,
> > > > about
> > > > > > kinematic analysis of over-determinate systems.
Unfortunately I
> > > > have
> > > > > > not been doing successful doing so yet. The reason is
the feet;
> > > > > > since there are only two markers here it creates
problems for
> > > > the
> > > > > > algorithm to determine primarily the eversion/inversion
dof, in
> > > > a
> > > > > > good way. Especially the left foot is problematic
because the
> > > > heel,
> > > > > > mallous and metatarsal markers are almost lying on
straight
> > > > line.
> > > > > > This problem will persist regardless of which method we
use to
> > > > > > impose the movement on the model. If you a have a
dataset with
> > > > more
> > > > > > markers on the forefoot it would be easier for the
algorithm to
> > > > > > solve this problem.
> > > > > >
> > > > > > You are right that improving the kinematics manually is
rather
> > > > > > difficult and time consuming; this was why I tried using
> > > > Michael’s
> > > > > > optimizer on your problem. If it had been successful the
output
> > > > > > would have been the optimized marker positions, segments
> > > > lengths,
> > > > > > and joint angles, but I am afraid that for the current
data set
> > > > this
> > > > > > is not possible, I had to abandon this strategy after
having
> > > > spend
> > > > > > quite some time to make it work.
> > > > > >
> > > > > > It is possible that scaling the left and right side
differently
> > > > will
> > > > > > improve this model. This can not be done straight away
and
> > > > requires
> > > > > > some changes to be made in the BRep.
> > > > > >
> > > > > > I will try list the needed changes here:
> > > > > > 1. In the AnyMan you need to make a left and right side
folder
> > > > > > obviously, containing the segment lengths for the two
sides.
> > > > > > 2. In the UniformScaling file you need to create a left
and
> > > > right
> > > > > > side scaling for the legs segments.
> > > > > > 3. In the BodyModels directory in the specific BodyModel
file
> > > > you
> > > > > > are using you need to introduce a new folder reference,
for
> > > > example
> > > > > > named ScalingRightLeft, in the Right and Left side
folders.
> This
> > > > > > reference should point at the scaling laws just created.
This
> > > > folder
> > > > > > reference should have the same name in both folders, but
point
> > > > at
> > > > > > different scalinglaws.
> > > > > > 4. In the Leg3d/Seg…any file you need to point at the
scaling
> > > > law
> > > > > > defined in the bodymodel right and left folder., instead
of the
> > > > > > existing scaling which are there today.
> > > > > >
> > > > > > This should introduce different scaling to the right and
left
> > > > side.
> > > > > >
> > > > > > I hope scaling the left and right side differently will
help
> > > > solve
> > > > > > the problem, but the inversion and eversion of the foot
will
> > > > > > continue to be a problem.
> > > > > >
> > > > > > Best regards
> > > > > > SÃ¸ren, AnyBody Support
> > > > > >
> > > > > >
> > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
wrote:
> > > > > > >
> > > > > > > Hi SÃ¸ren:
> > > > > > >
> > > > > > > I have improved the model a little bit. It can now run
to the
> > > > end.
> > > > > > > But, the kinematics is still a problem. The updated
model is
> > > > > > > Gait3D_rev3, which is up-loaded.
> > > > > > >
> > > > > > > John
> > > > > > >
> > > > > > > — In anyscript@yahoogroups.com, “johnzengwu” <jwu@>
wrote:
> > > > > > > >
> > > > > > > > Hi SÃ¸ren:
> > > > > > > >
> > > > > > > > Thanks for your time and efforts. I agree with you
that the
> > > > > > problem is
> > > > > > > > likely caused by the kinematics. But, it is a tough
job. I
> > > > have
> > > > > > spent
> > > > > > > > a lot of time to have my model scaled and motion
markers
> > > > > > adjusted.
> > > > > > > > When it is initialized, the model looks very good.
But, it
> > > > runs
> > > > > > only
> > > > > > > > 20 steps and stops. What I found is that the subject
in our
> > > > > > study is
> > > > > > > > unsymmetric. It is shown in my model, if you look at
it
> > > > closely.
> > > > > > The
> > > > > > > > thigh and shank lengths are different from left to
> right. We
> > > > also
> > > > > > > > found it in our direct physical measurements. This
may
> cause
> > > > the
> > > > > > > > problem. My questions is: how to adjust the left and
right
> > > > thigh
> > > > > > and
> > > > > > > > shank length separately?
> > > > > > > >
> > > > > > > > I have revised the MotionMarkerDrivers, but it does
not
> help.
> > > > > > > >
> > > > > > > > I have up-loaded my up-dated model “Gait3D-rev3”. All
> motion
> > > > > > markers,
> > > > > > > > which are not used to drive the model, have been
removed in
> > > > the
> > > > > > model
> > > > > > > > to make it clearere to view.
> > > > > > > >
> > > > > > > > Thanks,
> > > > > > > >
> > > > > > > > John
> > > > > > > >
> > > > > > > >
> > > > > > > > — In anyscript@yahoogroups.com, “AnyBody Support”
> > > > <support@>
> > > > > > wrote:
> > > > > > > > >
> > > > > > > > > Hi John
> > > > > > > > >
> > > > > > > > > I have looked a bit more into the model
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > I think the main problem could be the kinemtatics.
I
> > > > tried
> > > > > > again to
> > > > > > > > > insert a reaction force on the right and left
ankle
> > > > > > > > > inversion/eversion and the activity dropped from
15 to
> > > > 5.8.
> > > > > > This
> > > > > > > > > might be cause by the angle between the foot and
the floor
> > > > > > (tilt), it
> > > > > > > > > looks like he is walking on the lateral side of the
> feets.
> > > > > > > > >
> > > > > > > > > Secondly i looked at the motion of the pelvis, if
you
> look
> > > > the
> > > > > > > > > motion from the front there are some very large
rotations
> > > > of
> > > > > > the
> > > > > > > > > pelvis, and it can be seen that there is a large
error on
> > > > the
> > > > > > Asis
> > > > > > > > > markers. Therefore i tried adding reaction forces
to the
> > > > hip
> > > > > > > > > abduction and external rotatation together with
the ankle
> > > > > > inversion
> > > > > > > > > moment.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > AnyReacForce LeftSideJoints ={
> > > > > > > > >
> > > > > > > > > AnyKinMeasureOrg &ref2=
> > > > > > > > > Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > > > AnyKinMeasureOrg &ref5=
> > > > > > > > > Main.Model.HumanModel.Interface.Left.HipAbduction;
> > > > > > > > > AnyKinMeasureOrg &ref6=
> > > > > > > > >
Main.Model.HumanModel.Interface.Left.HipExternalRotation;
> > > > > > > > >
> > > > > > > > > };
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > AnyReacForce RightSideJoints ={
> > > > > > > > >
> > > > > > > > > AnyKinMeasureOrg &ref2=
> > > > > > > > >
Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > > > AnyKinMeasureOrg &ref5=
> > > > > > > > > Main.Model.HumanModel.Interface.Right.HipAbduction;
> > > > > > > > > AnyKinMeasureOrg &ref6=
> > > > > > > > >
Main.Model.HumanModel.Interface.Right.HipExternalRotation;
> > > > > > > > >
> > > > > > > > > };
> > > > > > > > >
> > > > > > > > > This helped the activity to drop to 1.4. I think
the
> > > > > > explanation is
> > > > > > > > > that the momentarms around the hip joint becomes
too
> small
> > > > for
> > > > > > this
> > > > > > > > > amount of pelvis “tilt”.
> > > > > > > > >
> > > > > > > > > So i think the next step will be to adjust the
motion.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > I looked at the moment curves and compared them to
the
> > > > curves
> > > > > > which
> > > > > > > > > can be found in book "Dynamics of human gait "
> > > > > > > > > http://www.kiboho.co.za/GaitCD/GaitBook.pdf
> > > > > > > > >
> > > > > > > > > On page 42 there is a curve for the moment around
the
> hip,
> > > > > > knee and
> > > > > > > > > ankle, when i compare the knee moment with the
> > > > > > > > > curve “Main.Study.Out_kinetics.M_Knee_Flex” it
looks
> quite
> > > > > > different?
> > > > > > > > >
> > > > > > > > > My impression is that it is the kinematics which
are
> > > > causing
> > > > > > these
> > > > > > > > > problems primarily. I think the scaling and the
marker
> > > > > > postions
> > > > > > > > > should be adjusted until the motion gives a
smaller
> > > > hip “tilt”
> > > > > > and
> > > > > > > > > the foot tilt are more horizontal and aligned with
the
> > > > floor.
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > Best regards
> > > > > > > > > SÃ¸ren, AnyBody support
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > — In anyscript@yahoogroups.com, “johnzengwu”
<jwu@>
> > > > wrote:
> > > > > > > > > >
> > > > > > > > > > Hi SÃ¸ren:
> > > > > > > > > >
> > > > > > > > > > Thanks for your suggestions. I have improved my
model,
> > > > as
> > > > > > > > > suggested.
> > > > > > > > > >
> > > > > > > > > > 1. The CenterOfPressure drivers have been
revised. Now,
> > > > I
> > > > > > drive
> > > > > > > > > > CenterOfPressure only in x-y plane, and raise it
to a
> > > > height
> > > > > > > > > of “pz”,
> > > > > > > > > > which can ben adjusted. By doing this, the
muscle force
> > > > in
> > > > > > > > > > gastrocnemius is reduced by 10%, still too high.
> (See my
> > > > > > revised
> > > > > > > > > > “enviroment.any” file.)
> > > > > > > > > >
> > > > > > > > > > 2. All section masses have been scaled. But this
has
> > > > little
> > > > > > > > > influence.
> > > > > > > > > > (see revised “man.any”)
> > > > > > > > > >
> > > > > > > > > > 3. I have checked the joint moment by
using “No_Muscle”
> > > > and
> > > > > > output
> > > > > > > > > the
> > > > > > > > > > joint muscle force. The sum of plus and minus
force
> > > > gives
> > > > > > the
> > > > > > > > > moment.
> > > > > > > > > > I have added a data processing unit
> > > > (“Output_kinetic.any”)
> > > > > > to
> > > > > > > > > output
> > > > > > > > > > the joint moment. Note: M and nM respresent
moment and
> > > > > > normalized
> > > > > > > > > > moment, respectively. When normalized to the
body mass,
> > > > the
> > > > > > joint
> > > > > > > > > > moments are consistent with published data.
> > > > > > > > > >
> > > > > > > > > > 4. I found that the soleus muscle seems not
active and
> > > > has
> > > > > > no force
> > > > > > > > > > output. All external forces are balanced by the
> > > > > > gastrocnemius
> > > > > > > > > muscles.
> > > > > > > > > > I have checked by varying the force magnitude in
> > > > > > environment.any.
> > > > > > > > > > Using “F=kforce(t)" with varied “k”, I am able
to
> > > > change
> > > > > > the
> > > > > > > > > ground
> > > > > > > > > > reaction force input. The output of the forces
of
> > > > > > gastrocnemius are
> > > > > > > > > > almost proportional to those in force plate.
> > > > > > > > > >
> > > > > > > > > > 5. In “JointsAndDrivers.any”, I have tested to
> switch on
> > > > > > > > > Reaction.Type
> > > > > > > > > > at HeelYLDrv and HeelYLDrv to
> Reaction.Type={Off,On,On},
> > > > the
> > > > > > soleus
> > > > > > > > > > forces then appear and gastrocnemius force
decrease by
> > > > 1/3.
> > > > > > It
> > > > > > > > > seems
> > > > > > > > > > the problem is at ankle joints.
> > > > > > > > > >
> > > > > > > > > > I appreciate it you can take time to look at my
revised
> > > > > > model, and
> > > > > > > > > > give me some further suggestions.
> > > > > > > > > >
> > > > > > > > > > The revised model “Gait3D-rev2” has been up-
loaded.
> > > > > > > > > >
> > > > > > > > > > John
> > > > > > > > > >
> > > > > > > > > > — In anyscript@yahoogroups.com, “AnyBody
Support”
> > > > > > <support@>
> > > > > > > > > wrote:
> > > > > > > > > > >
> > > > > > > > > > > Hi John
> > > > > > > > > > >
> > > > > > > > > > > I have tested your model and found a few things
> > > > > > > > > > >
> > > > > > > > > > > 1: the position of the CentrOfPressure segment
> seem to
> > > > be
> > > > > > too
> > > > > > > > > low.
> > > > > > > > > > > This can be corrected by adding a new node in
the
> > > > > > globalref
> > > > > > > > > which is
> > > > > > > > > > > located higher, and then refering to this node
in the
> > > > > > measures
> > > > > > > > > used
> > > > > > > > > > > for driving this segment.
> > > > > > > > > > >
> > > > > > > > > > > 2: The ankle eversion/inversion seems a bit
strange,
> > > > this
> > > > > > might
> > > > > > > > > > > cause problems for the muscles carrying this
dof. In
> > > > order
> > > > > > to
> > > > > > > > > see if
> > > > > > > > > > > this is a problem, i inserted these lines in
the
> > > > > > > > > > > JointsAndDrivers.any file
> > > > > > > > > > >
> > > > > > > > > > > AnyReacForce AnkleEversion ={ //stc added
these
> > > > > > reactions
> > > > > > > > > because
> > > > > > > > > > > AnyKinMeasureOrg
> > > > > > > > > > >
> > > > &ref1=Main.Model.HumanModel.Interface.Right.AnkleEversion;
> > > > > > > > > > > AnyKinMeasureOrg
> > > > > > > > > > >
> > > > &ref2=Main.Model.HumanModel.Interface.Left.AnkleEversion;
> > > > > > > > > > > };
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > The lines adds a reaction force to ankle
> > > > > > inversion/eversion
> > > > > > > > > dofs. It
> > > > > > > > > > > help the activity to drop a bit, but not
enough.
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > 3: Then i switched off the muscles and looked
at the
> > > > > > moments.
> > > > > > > > > The
> > > > > > > > > > > left knee moment looks a bit strange you can
see by
> > > > > > looking at
> > > > > > > > > the
> > > > > > > > > > > property
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > >
> > > > >
> "Main.Study.Output.Model.HumanModel.Left.Leg.JointMuscles.KneeJnt.d
> > > > > > > > > of
> > > > > > > > > > > 0.Muscle..Ft”
> > > > > > > > > > >
> > > > > > > > > > > it will give you two curves one of them is the
> > > > negative
> > > > > > moment
> > > > > > > > > the
> > > > > > > > > > > other one the positive moment. Normally the
negative
> > > > one
> > > > > > should
> > > > > > > > > be
> > > > > > > > > > > dominant in these models and it should not
change
> sign
> > > > > > during
> > > > > > > > > the
> > > > > > > > > > > stance phase as it does in this case, so i
think this
> > > > is
> > > > > > worth
> > > > > > > > > > > investigating a bit. I have not been able to
find the
> > > > > > cause of
> > > > > > > > > it,
> > > > > > > > > > > typical reasons could be kinematics of the
left leg,
> > > > the
> > > > > > way the
> > > > > > > > > > > force is applied etc.
> > > > > > > > > > >
> > > > > > > > > > > 4: The last thing i have noticed is about the
> scaling,
> > > > in
> > > > > > the
> > > > > > > > > > > AnyMan.any file it seems that your subject
have a
> > > > weigth
> > > > > > of 78
> > > > > > > > > kg,
> > > > > > > > > > > but looking at the segment masses of the thigh
and
> > > > shank
> > > > > > they
> > > > > > > > > are
> > > > > > > > > > > unchanged compared to the original Gait3D
model. I
> > > > think
> > > > > > these
> > > > > > > > > > > values should be slight higher.
> > > > > > > > > > >
> > > > > > > > > > > I hope it helps you move on, although it do
not solve
> > > > the
> > > > > > entire
> > > > > > > > > > > problem.
> > > > > > > > > > >
> > > > > > > > > > > Best regards
> > > > > > > > > > > SÃ¸ren, AnyBody support
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > — In anyscript@yahoogroups.com, “johnzengwu”
<jwu@>
> > > > > > wrote:
> > > > > > > > > > > >
> > > > > > > > > > > > Hi SÃ¸ren:
> > > > > > > > > > > >
> > > > > > > > > > > > I am applying Gait3D using my own data. The
model
> > > > seems
> > > > > > > > > working and
> > > > > > > > > > > > the magnitude of the joint moments are also
OK. But
> > > > the
> > > > > > > > > predicted
> > > > > > > > > > > > magnitude of muscle forces are very strange.
> > > > Especially,
> > > > > > the
> > > > > > > > > force
> > > > > > > > > > > in
> > > > > > > > > > > > gastrocnemius reaches over 4500 N, it’s too
large;
> > > > wile
> > > > > > force
> > > > > > > > > in
> > > > > > > > > > > > soleus is only 200-300N, too small. I have
checked
> > > > your
> > > > > > sample
> > > > > > > > > > > Gait3D,
> > > > > > > > > > > > the muscle forces are OK. I have run several
other
> > > > data
> > > > > > sets,
> > > > > > > > > and
> > > > > > > > > > > the
> > > > > > > > > > > > results are similar. It seems there is a
systematic
> > > > > > problem in
> > > > > > > > > my
> > > > > > > > > > > > model, but I cannot find it. Can you look at
my
> > > > model
> > > > > > and give
> > > > > > > > > me
> > > > > > > > > > > some
> > > > > > > > > > > > idea where the problem may be?
> > > > > > > > > > > >
> > > > > > > > > > > > I have up-loaded my model: “Gait3D-revised”
> > > > > > > > > > > >
> > > > > > > > > > > > Thanks,
> > > > > > > > > > > >
> > > > > > > > > > > > John Wu
> > > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>