RRA: High Residual Torques During Treadmill Walking

[Edward Syrett]

Hi all,
I'm trying to run RRA on some data that I gathered from walking on a split belt treadmill at 1.5 m/s, with the intent to run CMC after. I am still getting relatively high residual torques, even after adjusting tracking weights, re-scaling the model, and adjusting the actuator optimal forces. My main concern is the Mz residual torque, which max is frequently exceeding the recommended 50-75 N-m. And then, CMC fails because it is unable to generate the appropriate torques. Attached is a picture of the residual force output for one subject's RRA solution, around 79 kg. This is a second iteration of RRA, after adjusting the model, and is relatively representative of my data as a whole.
The Mz force seems to peak during each heelstrike on the treadmill. The residuals seem to indicate it is preventing a forward tipping of the model, making me think that this Mz torque is functioning to oppose the torque of the treadmill belt, which may not be accounted for in my simulation.
Has anyone run into this before and have any tips on how to reduce this? I checked the ground reaction force data, and even reimported it using force plate origin rather than COP with similar results. I have altered my IK solution, tried re-scaling, changed RRA task weights, and changed actuator torques with only mild differences in results. I feel like I am running out of levers to pull.
Am I trying to collect too many steps? This is 3 per side. Do I need to reduce to one at a time? I want to try to have at least 3 steps worth of data so I can average them for my results.
Also, just so I am sure, when I iterate RRA, I always use the IK solution as the input, correct? I have tried using the first RRA kinematics solution as the input for the second iteration of RRA, but it makes the residuals worse rather than better. Using the original IK solution for each RRA input seems to be the way to somewhat reduce the residuals.
I can provide any additional information as well. Thank you!

[Aaron Fox]

Hi Edward,

I don't have a solution for you - but thought I'd flag that I have also seen this 'drift' in residuals when simulating multiple steps. I never really figured out a solution, and logically I don't think it should happen - but I could run RRA and get increasing residuals, but then run the last step individually and have lower residuals on that exact same step. It sort of beat me and I just started focusing on individual steps - so as I said, no solution but you're not the only person to have these issues with RRA!

Aaron