Hi, I'm hoping some one can help me. I've been trying to run RRA on my model. However, it keeps failing on me. And if the analysis finishes, the output kinematics only have one or two states. I've tried everything I can think of. I've modified the kv, kp, weights, integration parameters, precision, etc. I've double checked to make sure the ground reaction forces and the kinematics match up. I am completely clueless as to what I'm doing wrong.
Can any one help me?
Thank you
Joseph Alston
RRA failure
-
Ayman Habib
- Posts: 2235
- Joined: Fri Apr 01, 2005 12:24 pm
RE: RRA failure
Hi Joseph,
Please check the trouble-shooting page for RRA/CMC at
http://wiki.simtk.org/opensim/rraCmcPage
In general (with correct setup and arbitrarily large residuals) you should be able to reproduce your motion/kinematics almost exactly. Then you should go back to see which residuals are large and why.
Hope this helps,
-Ayman
Please check the trouble-shooting page for RRA/CMC at
http://wiki.simtk.org/opensim/rraCmcPage
In general (with correct setup and arbitrarily large residuals) you should be able to reproduce your motion/kinematics almost exactly. Then you should go back to see which residuals are large and why.
Hope this helps,
-Ayman
-
Joseph Alston
- Posts: 26
- Joined: Tue May 13, 2008 9:51 am
RE: RRA failure
Ayman,
Thank you for the information. After I reviewed my setup files, I think that there are no error in the setup. However, the model doesn't track well at all. I've tried playing around with the kp and kv values, as well as the weights. However, I've so far been unable to have any success. Do you have any suggests about how to find kp, kv, weights, and strengths that will get my model to track properly?
Thank you,
Joseph
Thank you for the information. After I reviewed my setup files, I think that there are no error in the setup. However, the model doesn't track well at all. I've tried playing around with the kp and kv values, as well as the weights. However, I've so far been unable to have any success. Do you have any suggests about how to find kp, kv, weights, and strengths that will get my model to track properly?
Thank you,
Joseph
-
Ayman Habib
- Posts: 2235
- Joined: Fri Apr 01, 2005 12:24 pm
RE: RRA failure
Joseph,
Can you run inverse dynamics? Are the results reasonable? That may give you a clue.
Also, you could try isolating if it's a "controls" problem due to incorrect "control-constraints" by running for a short time and looking at partial results to see if some reserves do saturate incorrectly.
Checking the flag to "Ajust-Kinematics" gives the optimizer some freedom at the expense of tracking quality, the User Guide would have details about that.
Good Luck,
-Ayman
Can you run inverse dynamics? Are the results reasonable? That may give you a clue.
Also, you could try isolating if it's a "controls" problem due to incorrect "control-constraints" by running for a short time and looking at partial results to see if some reserves do saturate incorrectly.
Checking the flag to "Ajust-Kinematics" gives the optimizer some freedom at the expense of tracking quality, the User Guide would have details about that.
Good Luck,
-Ayman
-
Joseph Alston
- Posts: 26
- Joined: Tue May 13, 2008 9:51 am
RE: RRA failure
Ayman,
Thank you for your help. I just saw your message today. I'm going to try what you have suggested. I do have some questions though.
I can run inverse dynamics. Although, the inverse dynamics have what should be the residuals as being large numbers, around 2500. However, as I'm not realy interested in the lower body, those residuals might be fine.
How do I check to see if the forces are saturating?
When running RRA, the model kept failing. I tried increasing the optimal forces, with little success. When I looked at the Inverse Dynamic results, I relized that the idealized forces would need to act in both the negative and positive directions. And some of the forces would act primarily in the negative direction. How do the idealized forces deal with acting in both directions? Can they work in both directions? If the absolute magnitude of a force is greater in the negative direction than the positive direction, should I change the optimal force to a negative to reflect that? Also is it a problem if the optimal forces in the actuators file are much larger than they need to be?
Thank you again for your help,
Joseph
Thank you for your help. I just saw your message today. I'm going to try what you have suggested. I do have some questions though.
I can run inverse dynamics. Although, the inverse dynamics have what should be the residuals as being large numbers, around 2500. However, as I'm not realy interested in the lower body, those residuals might be fine.
How do I check to see if the forces are saturating?
When running RRA, the model kept failing. I tried increasing the optimal forces, with little success. When I looked at the Inverse Dynamic results, I relized that the idealized forces would need to act in both the negative and positive directions. And some of the forces would act primarily in the negative direction. How do the idealized forces deal with acting in both directions? Can they work in both directions? If the absolute magnitude of a force is greater in the negative direction than the positive direction, should I change the optimal force to a negative to reflect that? Also is it a problem if the optimal forces in the actuators file are much larger than they need to be?
Thank you again for your help,
Joseph
-
Joseph Alston
- Posts: 26
- Joined: Tue May 13, 2008 9:51 am
RE: RRA failure
Ayman,
I finally figured out one of the major problems that my model had. My model has a number of segments in it where I don't know the mass. I couldn't just leave the mass and inertial parameters at 0, so I put them at .0000001. However, this mass is too small for the CMC algorithm to handle properly in my model. I increased the mass and inertial parameters to .0001, an amount that is still insignificant, but large enough for the program. And now the model will not fail when running RRA.
The problem I have now is that tracking error in the joints that contain one of those segments with the insignificant masses keeps increasign with time. I think that it is a problem with the integration parameters, but I'm not sure. Has anyone else had a problem with a growing tracking error in RRA? If so, how did you fix it?
Thank you for your help,
Joseph
I finally figured out one of the major problems that my model had. My model has a number of segments in it where I don't know the mass. I couldn't just leave the mass and inertial parameters at 0, so I put them at .0000001. However, this mass is too small for the CMC algorithm to handle properly in my model. I increased the mass and inertial parameters to .0001, an amount that is still insignificant, but large enough for the program. And now the model will not fail when running RRA.
The problem I have now is that tracking error in the joints that contain one of those segments with the insignificant masses keeps increasign with time. I think that it is a problem with the integration parameters, but I'm not sure. Has anyone else had a problem with a growing tracking error in RRA? If so, how did you fix it?
Thank you for your help,
Joseph
-
Ayman Habib
- Posts: 2235
- Joined: Fri Apr 01, 2005 12:24 pm
RE: RRA failure
Joseph,
The reason idealized forces would not act in both directions would likely be due to an error in the "control_constraints" file. Please make sure that there are
<default_min> your_min_here </default_min>
<default_max> your_max_here </default_max>
specified in the control constraints file you're using (since the defaults at the top of the file are probably .02 to 1.0).
Good luck,
-Ayman
The reason idealized forces would not act in both directions would likely be due to an error in the "control_constraints" file. Please make sure that there are
<default_min> your_min_here </default_min>
<default_max> your_max_here </default_max>
specified in the control constraints file you're using (since the defaults at the top of the file are probably .02 to 1.0).
Good luck,
-Ayman
-
Joseph Alston
- Posts: 26
- Joined: Tue May 13, 2008 9:51 am
RE: RRA failure
Ayman,
That is good to know about the "constraints file", I double checked that, and my constrainst file has the min and max being -1 and 1, respectively. So thankfully I shouldn't be having any problems from that.
However, I some of my joints in my model are still progressively going out of control as I run RRA. The joints I question have the weights at 90, kp at 100 and kv at 20. The mass of these joints is also insignificant, .0001.
I'm not sure what the problem is. Do you have any suggestions?
Joseph
That is good to know about the "constraints file", I double checked that, and my constrainst file has the min and max being -1 and 1, respectively. So thankfully I shouldn't be having any problems from that.
However, I some of my joints in my model are still progressively going out of control as I run RRA. The joints I question have the weights at 90, kp at 100 and kv at 20. The mass of these joints is also insignificant, .0001.
I'm not sure what the problem is. Do you have any suggestions?
Joseph