- Our knee loads are slightly lower than the ones found experimentally. However, the study was only done on one person and the deviations are pretty high. Being within one standard deviation is already a good validation. However, the shape doesn't match. This could be due to the fact that our activation patterns and our handle force are piecewise constant. We end up with a much more constant knee load, compared to the experiment where it varies continuously.
- The model was created with muscle parameters taken from cadavers of elderly people, rather than data coming from MRI of young healthy people. This is how we tried to recreate the atrophy of the lower limbs' muscles of people suffering from a spinal cord injury. We also lowered the activations excitations to 0.3 and 0.6. It goes further in this direction, as well as takes into account that 2 electrodes on each face of the leg can't stimulate maximally all the fibers of all the muscles of the thigh. Further study on how much to lower the peak isometric force / the max activation should be done, as it has a direct effect on the magnitude of the knee loads.
Another axle of study could also investigate the effect of activating the calf muscles instead of the thigh muscles on knee loads and bone density.
Reproducing our results:
This file (readme.txt) explains how to use the different files to reproduce our results.
You can download our final model here.
The multiple iterations of the model are available here:
- Bodies + Joints: Stripping down 1018 model Building from scratch in C++
- Adding the Weld constraint: Stripping down 1018 model Building from scratch in C++
After validating that up to this part the 2 models were behaving similarly, we enhanced only the "stripped down" model in xml.
- Adding the 3 muscles
- Replacing the pin joint in the knee by a more accurate representation of the patella's role
- Replacing the hamstrings (bifem long head) by the bifem short head
The goal of this replacement was to solve the problem that the majority of the extension part was done when the hamstrings were activated. Having the short head muscle (which only crosses the knee) enabled us not to have to worry about the hip extension action of the long head. However, this muscle is 5 times weaker so it didn't function properly alone. This model was not giving satisfactory results.
- Final model: bifem long head + short head
We would like to thank Scott Delp and the entire ME 485 teaching team, especially to Calvin Kuo for all the help and guidance with our model.