simtk.org

Hello Everyone,

I'm trying to run forward dynamics for gait study .I have controls and state file after running RRA+CMC tool

Now the problem is that when i use this files in forward dynamic tool the model does some crazy movement. To overcome this problem i have tried to Append the force set (CMC Actuator.xml) where Coodinate Actuator Name have optimal force of 3 and all the Residual Actuators in the model have optimal force of 1, having min control (-inf) and max control (inf).
The Thelen muscle having maximum isometric force of 6000 N and optimal force is 1 .osim model (2392).

With every change in isometric force i can see the different movement the model after running the FD tool

Could please anyone suggest what I'm missing?

Regards,
Rahul

Hi Rahul,

If your CMC performed correctly/reasonable then going to ForwardDynamics is straightforward assuming you use the same settings (since CMC internally performs small forward integrations). There maybe numerical drift but that would happen late in integration, you just need to make sure to use the same initial-states, CMC Actuators/reserves/residuals and external-loads.

Hope this helps,
-Ayman

Thank you for your reply sir.

I have included Hunts crossely force contact in the gait model for simulation but while running the FD tool in force reporter I cannot see the contact force (i dont have GRF's).

I can see the desired movement of the model after running the FD tool

Can you tell me what is the problem with contact forces

Regards,
Rahul

Hi Rahul,

If you're doing gait without GRF than it's questionable what conclusions you can make running RRA/CMC. The premise of CMC is that you provide best measurements of external forces (GRF etc.) and then RRA/CMC account for small errors and compute muscle activations to reproduce the motion. Without these you can still get CMC to run employing residuals/reserves but the results are hard to interpret.

That aside, running CMC with contact is challenging for many other reasons e.g. Large force introduced at contact that muscles can't account for fast enough, drifts etc.

It would be important to understand what's not working in your contact force but it would help a lot if you can isolate that into a small simple model and report a bug/issue if it's not working as expected.

Hope this helps,
-Ayman

Thank your for your prompt response sir.

My study mainly uses forward dynamic approach because it is impossible for me to collect GRF's experimentally. To understand the muscle excitation of the movement i have to execute CMC tool. This muscle excitation I can use in FD tool with contact model to generate the contact forces if I'm not wrong ? . Also, I have used same set of contact sphere as in optimal jump example for gait simulation.

If somehow i manage to get GRF's and then run CMC to get control and states file. Do you think it would be necessary for me to use the GRF's in FD tool to get the same movement or contact modelling can take care of that.

I have some problem in using CMC tool.For my case I am simulating the model's walk for the duration of 1 hour.I want to get the force as the output.When I use CMC tool, its taking around 24 hours for 2 minutes computation.So,if i want to compute for 3600 seconds then it may take 30 days approx.How to do it at less expense of time ?.Is there any other way to find force of muscle for 3600 seconds simulation at 2 hours time ?.
Please help me out.....

There several options you might consider, depending on your research question and potentially in combination:
1. Sample your data: generate simulations of a small number of representative gait cycles for each hour.
2. Use the Static Optimization tool instead (https://simtk-confluence.stanford.edu:8 ... timization).
3. Run simulations in parallel on multiple CPU threads, a computing cluster, etc.
4. Use a simpler model that simulates faster.