I am running CMC using outputs from RRA and optimized tracking weights for running at 3.5 m/s (both people wearing prostheses and people without an amputation).
I have consistently had trouble getting CMC to initialize when using the RRA kinematics - the process takes around 45 mins to 1 hour to start computing.
The only consistent difference I have noticed is that RRA will occasionally compute a datapoint that is unequally spaced with the rest. For example, the simulation will have timesteps of 7.898, 7.8987, 7.899 instead of 7.898, 7.899. I generally receive an error reading:
Storage.resample: WARNING: resampling at time step 1.159e-005 (but minimum time step is 1.034e-005)
when I try to use CMC with the RRA output kinematics, making me think that the RRA timestep is too small to use with CMC. If I manually remove these data points (or use an equally spaced kinematics file) the simulation will initialize in barely no time.
Is there any reason why RRA computes these intermediate points, or why the resampling takes so long in the CMC process? Additionally, can I avoid this somehow? Of course, manually resampling the RRA results would be an option, but I'd prefer to get to the bottom of this issue first.
Thanks!
Lauren
CMC Slow to initialize using RRA kinematics
-
Lauren Sepp
- Posts: 3
- Joined: Mon Sep 02, 2013 4:10 pm
-
Dimitar Stanev
- Posts: 1096
- Joined: Fri Jan 31, 2014 5:14 am
Re: CMC Slow to initialize using RRA kinematics
It is possible that you are analyzing a long duration movement. Try limiting the analysis to 1-2 seconds, depending on the movement. Furthermore, check the time step of the input data. It is possible that the sampling is very dense and this may cause delays in the filtering of the signal.
-
Lauren Sepp
- Posts: 3
- Joined: Mon Sep 02, 2013 4:10 pm
Re: CMC Slow to initialize using RRA kinematics
Hi Dimitar,
Great suggestions - I've tried many different time periods and locations of start ranging from about 0.3 seconds to 1.5 seconds. The input data to RRA is all at 100Hz - not uncommon (and might I add, a little sparse) for running/sprinting data.
Do you think that this issue could be related to the RRA algorithm parameters? Too many number of steps or step size in the integrator? I've played around a little bit with that, but can't find anything repeatable.
Thanks!
Lauren
Great suggestions - I've tried many different time periods and locations of start ranging from about 0.3 seconds to 1.5 seconds. The input data to RRA is all at 100Hz - not uncommon (and might I add, a little sparse) for running/sprinting data.
Do you think that this issue could be related to the RRA algorithm parameters? Too many number of steps or step size in the integrator? I've played around a little bit with that, but can't find anything repeatable.
Thanks!
Lauren