Arnim Henze

04-28-1997, 07:12 PM

Dear listmembers,

A few weeks ago I posted a question concerning invers dynamics with

the multibody simulation system DADS.

I was wondering whether any one else had the same problems with

DADS's cubic-splineinterpolation for the inputdata for invers dynamical

calculations which resulted in zig-zag joint-torques and -forces.

I did not get too many replies and I am still interested in other

suggestions and experiences. Here is a summery of the answers I got:

Brian T. Fay, Technology Consultant form Jacksonville, IL wrote,

"... it sounds like you may have noise problems ...

The NEwton-Raphson interpolation used by DADS ... is described in the

textbook written by Haugh. Unfortunately, I believe the book may be

out-of-print. I have a copy and could forward the ISBN to you if you

like, or you could try to contact Haugh who is at the University of Iowa.

Also, CADSI has been very helpful in the past when I have contacted them."

We surely know, that these are not noise problems. We have a copy of

E.Haugh's textbook, but we can't find any solution for the problem there.

Paul Bourassa advised us to

"... to give a look at the book "computer-aided analysis of

mechanical Systems" by Parviz E. Nikravesh Prentice-Hall ISBN-0-13-164220-0.

The book is out of print but you might get it through library loans. Nik

phone 602-621-2235 He has an Email but I have lost it.

Nikravesh was working with the group of Ed Haug in 1983-84 and wrote a

great deal of the DADS code. In his book, he gives a source code which I

think must be quite similar to DADS. You may also want to contact

Nikravesh who is a professor at the university of Arizona. You might get

some more code from him. ... The problem with DADS is often that you do not

know what is going around because you do not have the source. For example,

what is the algorithm that is being used for integration. ..."

This might be helpful - as long as the special implementation of the

inverse dynamics tool in DADS doesn't lead to these problems. We'll try

to get Parviz E. Nikravesh's book here in Germany or try to contact him.

Gail Jeffries from Brunel University wrote,

that he uses the Mechanical Analysis package called ADAMs to calculate

joint torques and he has the same problem with ADAMs.

"The only solution I can suggest, and I'm about to try this myself, is a two

stage approach. I.e., step one use the displacement data to calculate the

velocities, step two, use the calculated velocity to drive your model, and fit

the spline to -this- data."

I'm not sure whether I understand what you mean, but we tried something

similar and it worked fairly well - but, since it is a work around, it's

uncomfortable. We numerically calculated velocities (or accelerations) from

the displacement data, fitted a cubic spline to those values, numerically

integrated the interpolated data once (or twice) and fed the result into

DADS. This procedure should lead to some kind of a spline interpolation of

higher than 3rd. order.

Dr. Ulrich Glitsch from German Sport University Cologne wrote,

that he has been working with DADS for 6 years, and that he is satisfied

with the system, though he has had the same problems. The zig-zag torques and

forces are numerically correct - errors are minimized at the input data points

and direct dynamics with the calculated torques and forces leads to correct

kinematics. DADS does not take the variations of forces and torques in time

into account, no smoothing is done. The more dense the input values, the

smoother are calculated torques and forces. The curves are more pretty then, but

not necessarily more correct, which is a general problem of inverse dynamics and

does not really have anything to do with DADS. He does inverse dynamics with

handmade software which in his opinion is much more effective in the end than to

do it with DADS.

I think, your answer brought some light into our problem, thanks a lot.

Thanks to everyone who answered.

Arnim Henze.

--

-------------------------------------------------------------------------

Arnim Henze

Institut f\"ur

Astronomie und Astrophysik

- Abteilung Biomechanik - email: henze@tat.physik.uni-tuebingen.de

Auf der Morgenstelle 10 Tel.: 07071-29 78654

D-72076 T\"ubingen Fax : 07071-29 5889

-------------------------------------------------------------------------

A few weeks ago I posted a question concerning invers dynamics with

the multibody simulation system DADS.

I was wondering whether any one else had the same problems with

DADS's cubic-splineinterpolation for the inputdata for invers dynamical

calculations which resulted in zig-zag joint-torques and -forces.

I did not get too many replies and I am still interested in other

suggestions and experiences. Here is a summery of the answers I got:

Brian T. Fay, Technology Consultant form Jacksonville, IL wrote,

"... it sounds like you may have noise problems ...

The NEwton-Raphson interpolation used by DADS ... is described in the

textbook written by Haugh. Unfortunately, I believe the book may be

out-of-print. I have a copy and could forward the ISBN to you if you

like, or you could try to contact Haugh who is at the University of Iowa.

Also, CADSI has been very helpful in the past when I have contacted them."

We surely know, that these are not noise problems. We have a copy of

E.Haugh's textbook, but we can't find any solution for the problem there.

Paul Bourassa advised us to

"... to give a look at the book "computer-aided analysis of

mechanical Systems" by Parviz E. Nikravesh Prentice-Hall ISBN-0-13-164220-0.

The book is out of print but you might get it through library loans. Nik

phone 602-621-2235 He has an Email but I have lost it.

Nikravesh was working with the group of Ed Haug in 1983-84 and wrote a

great deal of the DADS code. In his book, he gives a source code which I

think must be quite similar to DADS. You may also want to contact

Nikravesh who is a professor at the university of Arizona. You might get

some more code from him. ... The problem with DADS is often that you do not

know what is going around because you do not have the source. For example,

what is the algorithm that is being used for integration. ..."

This might be helpful - as long as the special implementation of the

inverse dynamics tool in DADS doesn't lead to these problems. We'll try

to get Parviz E. Nikravesh's book here in Germany or try to contact him.

Gail Jeffries from Brunel University wrote,

that he uses the Mechanical Analysis package called ADAMs to calculate

joint torques and he has the same problem with ADAMs.

"The only solution I can suggest, and I'm about to try this myself, is a two

stage approach. I.e., step one use the displacement data to calculate the

velocities, step two, use the calculated velocity to drive your model, and fit

the spline to -this- data."

I'm not sure whether I understand what you mean, but we tried something

similar and it worked fairly well - but, since it is a work around, it's

uncomfortable. We numerically calculated velocities (or accelerations) from

the displacement data, fitted a cubic spline to those values, numerically

integrated the interpolated data once (or twice) and fed the result into

DADS. This procedure should lead to some kind of a spline interpolation of

higher than 3rd. order.

Dr. Ulrich Glitsch from German Sport University Cologne wrote,

that he has been working with DADS for 6 years, and that he is satisfied

with the system, though he has had the same problems. The zig-zag torques and

forces are numerically correct - errors are minimized at the input data points

and direct dynamics with the calculated torques and forces leads to correct

kinematics. DADS does not take the variations of forces and torques in time

into account, no smoothing is done. The more dense the input values, the

smoother are calculated torques and forces. The curves are more pretty then, but

not necessarily more correct, which is a general problem of inverse dynamics and

does not really have anything to do with DADS. He does inverse dynamics with

handmade software which in his opinion is much more effective in the end than to

do it with DADS.

I think, your answer brought some light into our problem, thanks a lot.

Thanks to everyone who answered.

Arnim Henze.

--

-------------------------------------------------------------------------

Arnim Henze

Institut f\"ur

Astronomie und Astrophysik

- Abteilung Biomechanik - email: henze@tat.physik.uni-tuebingen.de

Auf der Morgenstelle 10 Tel.: 07071-29 78654

D-72076 T\"ubingen Fax : 07071-29 5889

-------------------------------------------------------------------------