RRA crashes using muscle-free models

[Andreas Christou]

Hi everyone,

I've been trying to use RRA in a case where I have an exoskeleton model attached to a skeletal model (without muscles) using bushing forces and RRA has been crashing. This was a case where I wanted to perform an ID analysis to calculate the forces applied by the human, to achieve a specific trajectory of an exoskeleton, without explicit knowledge of the kinematics of the human, and without knowledge of the bushing forces. I finally managed to get it to work so I thought I'd post these details (that I struggled to find online) to hopefully make someone's life easier in the future and get feedback from others on whether this approach can be improved or performed using a different tool.

1) Muscles: It seems that RRA requires at least 1 muscle to be present in the model to prevent it from crashing. So, if the goal is to calculate joint torques either coming from the human or an attached device (and a muscle-actuated model is not considered), it is faster to remove all the muscles from the model except one, and disable that one muscle from applying any force during the RRA using either the GUI or the 'model.osim' file.
2) Bushing forces: One way to allow for bushing forces to be considered during RRA, is to add RRA's actuators to your model without replacing the model's force set using the RRA tool or by modifying the 'RRA_Setup.xml' file in order to disable the <replace_force_set> element.
3) Initial conditions: A kinematics '.mot' file is required for running the RRA tool along with a tracking tasks file ('RRA_Tasks.xml'). If there are any coordinates whose trajectory is unknown, then the RRA tool can provide an estimate of the trajectory. To do this, the initial conditions need to be defined in the kinematics file (and the rest of the time frames can be filled with any value) and those coordinates need to be disabled in the tracking tasks file. This allows for the initial conditions to be met. Given the effect of a minor mismatch between two coordinates on stiff bushing forces, it will be important to get the initial conditions right to avoid high initial bushing forces.

In short, given the example files for Gait2354_Simbody, to perform a similar analysis, it was necessary to modify the kinematic '.mot' file (to get the initial conditions right), the 'RRA_Actuators.xml' file (to add actuators for the exoskeleton's degrees of freedom), the 'RRA_Tasks.xml' file (to define which DOFs need to be tracked), and the 'model.osim' file (obviously to add the exoskeleton bodies and joints, and disable the muscle forces).

Best,
Andreas