RRA Unexpected Results
- Mihir Patel
- Posts: 2
- Joined: Sun Aug 23, 2020 6:28 pm
RRA Unexpected Results
I've got an output for RRA, but the model is drifting and falling backwards. Attached are the files I used, any help would be appreciated!
- Attachments
-
- Pilot_NA_IK_results_2.mot
- (111.51 KiB) Downloaded 26 times
-
- Pilot_NA_02_grf.xml
- (2.45 KiB) Downloaded 26 times
-
- Pilot_NA_02_grf.mot
- (271.29 KiB) Downloaded 21 times
-
- gait2354_RRA_Tasks.xml
- (56.11 KiB) Downloaded 27 times
-
- gait2354_RRA_Actuators.xml
- (13.18 KiB) Downloaded 21 times
Tags:
- Mihir Patel
- Posts: 2
- Joined: Sun Aug 23, 2020 6:28 pm
Re: RRA Unexpected Results
This is a picture of the resulting RRA output.
- Attachments
-
- RRA Output.png (167.07 KiB) Viewed 317 times
- Carmichael Ong
- Posts: 401
- Joined: Fri Feb 24, 2012 11:50 am
Re: RRA Unexpected Results
Usually this means there's some discrepancy between the measured kinematics and external forces applied (that the forces don't quite satisfy the observed coordinate accelerations). I took a quick peek at how the GRFs look and saw some possible issues:
Example at left foot push-off: Example at tight foot push-off: Notice how these are large GRFs that aren't under the foot, which would tend to make your model rotate backwards. You may want to double check that your CoP locations are reasonable, and if they are not (this can happen especially at low forces, since you divide by the force to calculate CoP) you could zero out some of the forces when they are pretty small (maybe <10 N or < 50 N) and the CoP location doesn't look quite right.
One other question is: why didn't RRA track the kinematics anyway? This might be due to the low "optimal_force" for the TorqueActuators MX/MY/MZ. If they were increased, they would be "cheaper" for the optimization to use, however, you would then just see higher MX/MY/MZ values when you validate the results which would mean that the simulations need to be refined further to have better confidence in the results.
Example at left foot push-off: Example at tight foot push-off: Notice how these are large GRFs that aren't under the foot, which would tend to make your model rotate backwards. You may want to double check that your CoP locations are reasonable, and if they are not (this can happen especially at low forces, since you divide by the force to calculate CoP) you could zero out some of the forces when they are pretty small (maybe <10 N or < 50 N) and the CoP location doesn't look quite right.
One other question is: why didn't RRA track the kinematics anyway? This might be due to the low "optimal_force" for the TorqueActuators MX/MY/MZ. If they were increased, they would be "cheaper" for the optimization to use, however, you would then just see higher MX/MY/MZ values when you validate the results which would mean that the simulations need to be refined further to have better confidence in the results.
- Carmichael Ong
- Posts: 401
- Joined: Fri Feb 24, 2012 11:50 am
Re: RRA Unexpected Results
Oh I should also note that since I didn't have the scaled model, I used the generic gait2392 model, so if you overlay the forces it may look different for your case. How I got the GRF to show up along with the kinematics:
1. Load gait2392
2. Load the .mot file you provided from the IK results.
3. Under the corresponding Motion that you loaded (under the "Motions" drop-down), right-click the corresponding Motion and select "Associate Motion Data.." and select the "*_grf.mot" file provided in the files
1. Load gait2392
2. Load the .mot file you provided from the IK results.
3. Under the corresponding Motion that you loaded (under the "Motions" drop-down), right-click the corresponding Motion and select "Associate Motion Data.." and select the "*_grf.mot" file provided in the files