I’m using the AnyBody version 5.3 with the rep 1.5. I’m using the GaitUniMiami model. I see that the arms are not included in the model, but what about the ground reaction forces ? They account for the fact that the whole body is acting against the floor, including the arms (10 kg of additional mass). Thus, we should also simulate the arm in that model, right ?

I saw that the markers that simulated the arms were commented out. Why ?

If I put them back and include the right and left arms, the optimization algorithm is not working, showing me the following error message (with a lot of violated constraints):

ERROR(OBJ.MCH.KIN3) : C:/U…s/P…r/D…s/ETS/A…y/R…s/A…5/A…n/M…L/G…D/GaitUniMiamiTD.main.any : MotionOptimization.InitialConditions : Kinematic analysis failed in time step 0

I guess that I’m doing something wrong or I forgot to perform some modifications to the model to simulate the whole body kinematics.

I was able to make it work, using twice the number of steps in the optimisation process. Also, I had to add the scapular rhythms, but only in the optimization process, not in the inverse dynamic analysis. Otherwise, my model was overconstrained (by 6 constraints) in the inverse dynamic analysis.

The optimization is longer than without the arms, which I guess is normal. Also, the upper body is now bigger compared to the lower body. I know that this is caused by the optimization, which change anthropometry to minimize kinematic error between internal and external markers. My question is : Is there a limit of anthropometry scaling (ie. 10 % bigger or smaller) defined by the optimization process to avoid unrealistic optimized anthropometry ?

no, there is no limit for Anthropometric optimization on the scaling. It’s either on or off. The UniMiami model in AMMRv1.5 was never setup for foll body simulations including the arms, therefore the initial lengths are not very precise, also maybe some marker positions might not be ideal.
In the latest version of the repository we added a UniMiami full body model. Please have a look there.

Ok great. Do you agree that the ground reaction force is not the right one if we do not consider the arms in the model ? At least, we should adjust the ground reaction force to account for the absence of the arms.

Yes, but it depends. If you look at spine, you should change something. If you look into lower extremity joints, it might be ok. The residual force between pelvis and Sacrum will carry the difference …

Ok I will think about it. Here is another observation that worth a reflexion:

When I perform a simulation with the two GaitUniMiami models (ie. rep 1.5 with AMS 5.3 and rep 1.6.1 with AMS 6.0), I obtain different estimations of optimized pamareters from the kinematic optimization and different estimations of knee contact forces from the inverse dynamic analysis.

I know that it might be hard to answer, but which of these two versions of the GaitUniMiami is the most physiological and why.

Looking at both models, I would suggest to use the old one for now. In the new MoCap model we have changed, as Daniel described some marker positions and weights. This improved the standard Gait model, but it seems that this Miami model got worse.

We will most likely make an update for the Miami model soon.