Using RRA on arm26 can't work?

[Gary Lu]

Dear all, i am using arm26 model to go through each step of opensim's function including scaling, IK ,ID, SO and I'm now working on RRA.
However I finished the setupfile and others documents like RRA_task, actuators and ControlConstraint, but when I press start, the overall adjustments are exactly 0 for mass recommendation and mass center adjustment even when I apply the wrong reaction forces. I think the problem might appears in the force file. Here are just my thought.
When I used Lagrange's equation to find the needed force for adjusting mass properties, the reaction force in the shoulders joint in required, so I put the force components into the force file and the position I set is the origin of the shoulder's frame coordinates. however, the RRA result turn out to be 0 no matter the wrong or right force data. Can anyone tell me how to use RRA on arm26 or there's no way to use RRA on the model.

[Ton van den Bogert]

RRA is only applicable to underactuated models, where there are certain generalized forces/moments that are supposed to be zero. For instance, in gait models, external forces and moments associated with moving the pelvis in space are known to be zero. But if you calculate them from the movement and the equations of motion, they are not exactly zero. These are the residuals that must be reduced by altering the movement and/or model to make them consistent with each other.

arm26 is fully actuated, which means that any movement that you measure can be exactly reproduced by the simulation. This means that RRA is not needed.

Ton van den Bogert

[Gary Lu]

arm26 is fully actuated, which means that any movement that you measure can be exactly reproduced by the simulation. This means that RRA is not needed.

Thanks for replying, sorry I forgot to mentioned that I applied a external load on the forearm pretended it is carrying a bucket with water. So In the equations of motion we have to know the joint reactions in order to adjust the mass properties, am I right?

[Ton van den Bogert]

The external load does not matter, the model is still fully actuated. No matter what the motion and external load is, you can find generalized forces T that can simulate the motion exactly. These generalized forces (i.e. the two joint torques) could be solved from the equation of motion:

M*qdd + C(q,qd)*qd = J'*Fext + T

The only thing that can make the simulation infeasible is when the muscles are not strong enough to produce the required torques, but that is not related to RRA.

A 3D gait model is underactuated, the first six elements of T must be zero (the six generalized forces associated with spatial pelvis motion), and then you have N equations with N-6 unknowns. The equations can't be satisfied exactly until you do RRA. Well, almost. There used to be a method called REA (residual elimination) but reduction was found to be sufficient. Anyway, none of that is relevant to fully actuated models such as arm26.

Ton

[Gary Lu]

OK, I think i got it. thanks Ton.