Tweet
Dear Subscribers:
I would like to thank those who responded my inquiry: Dan Moran, Mehran
Armand, Tasos Karakostas, and Ben Fregly. Here is a summary of the
responses.
Thank you all for your attention,
Xudong Zhang
Center for Ergonomics
Dept of Industrial & Operations Engineering
The University of Michigan
----------------------------------------------------------------------------
------------------
One way to correct for these errors is to recalculate the segmental
angles for the proximal segment of the joint your analyzing based on the joint
angles calculated for all the joints leading up to the current joint. For
instance, if you've calculated hip and knee angles already and are trying to
calculate ankle angles, I would recalculate the shank's segmental angles from a
rotation matrix built up from hip and knee angles. By using the recalculated
shank angles and the measured foot angles, the orientation of the foot will be
more accurate in the overall model. Note, the ankle joint angles will probably
be less accurate than if you used both measured segmental angles but since your
interested simulation it is more important to get the segments aligned in the
global reference frame than to get accurate joint angles.
Daniel W. Moran, Ph.D. phone: 619-626-2133
The Neurosciences Institute fax: 619-626-2199
10640 John J. Hopkins Drive email: dmoran@nsi.edu
San Diego, CA 92121
----------------------------------------------------------------------------
------------------
To solve such problem as you noted, I optimized for the cartezian location
of the most distal point on my linkage at each time step. The optimization
objective was to find the joint angles that require minimal change to
correct for the position of the most distal link. Any optimization routine
can be used and the convergence at each time step was very fast
(within 15 to 50 iterations for each time step) I used BFGS
optimization method (this a a secant method that you can find it in the
function "fminu" of matlab in the optimization tool box.
Mehran Armand
Department of Mechanical Engineering
University of Waterloo, Waterloo, Ontario, Canada N2L-3G1
Phone: (519) 888-4567 ext. 2601
Fax : (519) 746-6776
email: mehran@nclab.uwaterloo.ca
----------------------------------------------------------------------------
------------------
look at the Ph.D. dissertation by Moran, Daniel William. The title is
"Effects of Orthopaedic Surgery on Musculoskeletal Control Strategies in
Children with Cerebral Palsy" available from UMI dissertation services.
Tasos Karakostas
The Ohio State University Hospitals
Gait Analysis Laboratory
1054 Dodd Hall
471 Dodd Drive
Columbus, Ohio 43210-1290
e-mail: Tasos@gait1.gait.ohio-state.edu
tel.: (614)293-4832
fax.: (614)293-4834
----------------------------------------------------------------------------
------------------
I would introduce some additional measurements to reduce your errors.
For example, if you measured the x-y coordinates of the end point of the
kinematic chain as well, then you could perform some kind of
least-squares analysis to distribute the errors in the other x-y coordinates of
interest.
Errors in the measured lengths of the segments will also introduce
errors into the x-y coordinates calculated from segment angle measurements. I
will be presenting a paper at the upcoming ASME Bioengineering conference
describing one approach for improving this problem for analyses of
pedaling -replace a length measurement with a well chosen angle
measurement. This
general approach has application for other systems as well - always
measure experimentally the most error-sensitive angles or positions in your
model (e.g., the end point position in an open-chain system). This should
provide better overall kinematic results than if these error-sensitive angles or
positions were calculated from other measurements.
Benjamin J. Fregly, Ph.D.
Research and Development Engineer
Pro/MECHANICA MOTION
Parametric Technology Corporation
2590 North First Street, Suite 200
San Jose, CA 95131
Tel: 408/953-8650
E-mail: bj@sj.ptc.com
----------------------------------------------------------------------------
------------------
I would like to thank those who responded my inquiry: Dan Moran, Mehran
Armand, Tasos Karakostas, and Ben Fregly. Here is a summary of the
responses.
Thank you all for your attention,
Xudong Zhang
Center for Ergonomics
Dept of Industrial & Operations Engineering
The University of Michigan
----------------------------------------------------------------------------
------------------
One way to correct for these errors is to recalculate the segmental
angles for the proximal segment of the joint your analyzing based on the joint
angles calculated for all the joints leading up to the current joint. For
instance, if you've calculated hip and knee angles already and are trying to
calculate ankle angles, I would recalculate the shank's segmental angles from a
rotation matrix built up from hip and knee angles. By using the recalculated
shank angles and the measured foot angles, the orientation of the foot will be
more accurate in the overall model. Note, the ankle joint angles will probably
be less accurate than if you used both measured segmental angles but since your
interested simulation it is more important to get the segments aligned in the
global reference frame than to get accurate joint angles.
Daniel W. Moran, Ph.D. phone: 619-626-2133
The Neurosciences Institute fax: 619-626-2199
10640 John J. Hopkins Drive email: dmoran@nsi.edu
San Diego, CA 92121
----------------------------------------------------------------------------
------------------
To solve such problem as you noted, I optimized for the cartezian location
of the most distal point on my linkage at each time step. The optimization
objective was to find the joint angles that require minimal change to
correct for the position of the most distal link. Any optimization routine
can be used and the convergence at each time step was very fast
(within 15 to 50 iterations for each time step) I used BFGS
optimization method (this a a secant method that you can find it in the
function "fminu" of matlab in the optimization tool box.
Mehran Armand
Department of Mechanical Engineering
University of Waterloo, Waterloo, Ontario, Canada N2L-3G1
Phone: (519) 888-4567 ext. 2601
Fax : (519) 746-6776
email: mehran@nclab.uwaterloo.ca
----------------------------------------------------------------------------
------------------
look at the Ph.D. dissertation by Moran, Daniel William. The title is
"Effects of Orthopaedic Surgery on Musculoskeletal Control Strategies in
Children with Cerebral Palsy" available from UMI dissertation services.
Tasos Karakostas
The Ohio State University Hospitals
Gait Analysis Laboratory
1054 Dodd Hall
471 Dodd Drive
Columbus, Ohio 43210-1290
e-mail: Tasos@gait1.gait.ohio-state.edu
tel.: (614)293-4832
fax.: (614)293-4834
----------------------------------------------------------------------------
------------------
I would introduce some additional measurements to reduce your errors.
For example, if you measured the x-y coordinates of the end point of the
kinematic chain as well, then you could perform some kind of
least-squares analysis to distribute the errors in the other x-y coordinates of
interest.
Errors in the measured lengths of the segments will also introduce
errors into the x-y coordinates calculated from segment angle measurements. I
will be presenting a paper at the upcoming ASME Bioengineering conference
describing one approach for improving this problem for analyses of
pedaling -replace a length measurement with a well chosen angle
measurement. This
general approach has application for other systems as well - always
measure experimentally the most error-sensitive angles or positions in your
model (e.g., the end point position in an open-chain system). This should
provide better overall kinematic results than if these error-sensitive angles or
positions were calculated from other measurements.
Benjamin J. Fregly, Ph.D.
Research and Development Engineer
Pro/MECHANICA MOTION
Parametric Technology Corporation
2590 North First Street, Suite 200
San Jose, CA 95131
Tel: 408/953-8650
E-mail: bj@sj.ptc.com
----------------------------------------------------------------------------
------------------