Multiple wrapping surfaces on one MTU in double pendulum

[Jack Withers]

Hello, I'm having trouble with the behavior of wrapping surfaces and hope someone can help me understand why.

The model I wish to create is a pendulum with multiple arms, a cylindrical wrapping surface at each of the pendulum's joints, and one single MTU that originates from the top most arm of the pendulum and inserts into the last arm. I am trying to add each of the joint wrapping surfaces to the MTU so that when the pendulum bends away from MTU, it wraps around the entirety of the pendulum smoothly.

However, it appears that when the MTU attempts to wrap around multiple wrapping surfaces at once, its wrapping behavior becomes erratic.

I've constructed a simplified version of the model to illustrate. Here is what the model looks like when the coordinate is at -8 degrees:

Screen Shot 2022-10-17 at 3.07.41 PM.png (73.76 KiB) Viewed 288 times

And when the coordinate is bent slightly more to -9 degrees:

Screen Shot 2022-10-17 at 3.07.58 PM.png (70.13 KiB) Viewed 288 times

The expected behavior is the MTU continues to smoothly wrap around the wrapping surfaces as the pendulum bends. Instead there is strange and complex wrapping behavior once the MTU starts to contact multiple wrapping surfaces at once. We've confirmed this is not a visualization issue by looking at the muscle tendon length plot and the moment arm plots in OpenSim.

As a final note, it seems that changing the wrapping surface quadrant from "+x" to "all" fixes the erratic wrapping behavior, but functionally the wrapping surfaces still don't behave how we want in this case. We'd like to get it working properly with the +x quadrants.

I've attached the model for reference. I would greatly appreciate any insight regarding this issue.

[Thomas Uchida]

The example model is very helpful. It seems that the issue may be occurring when the muscle is only barely touching one of the wrap surfaces. For example, if you move wrap3 and wrap4 away from link2 (e.g., setting wrap3 translation to [0.1, 0.25, 0] and wrap4 translation to [0.1, 0, 0]), then the muscle never grazes past wrap3 and it wraps as expected. If this is just for a toy example, perhaps a solution like this would suffice (changing the locations and/or radii of the cylinders); otherwise, some other modifications may need to be investigated to capture the desired functionality while avoiding numerical issues in the algorithm.

[Jack Withers]

Thanks Thomas. The model I posted was a very simplified version of what we are hoping to do, in that the previous model I attached only had one wrapping surface at a joint. Here's a model that is a bit more like what we ultimately want to achieve, with all wrapping surfaces at joints.

Screen Shot 2022-10-18 at 12.25.30 PM.png (78.44 KiB) Viewed 134 times

Screen Shot 2022-10-18 at 12.27.14 PM.png (90.46 KiB) Viewed 134 times

I'm not sure if this will give you any more insight, just thought I'd post it in case. We're trying to model a mammal tail based on anatomical data so there isn't much wiggle room with how we can change the wrapping surfaces. If the issue doesn't seem like a simple fix I think we're going to try a via point approach.