Sorry it's taken me so long to post the videos of my tests. A couple of editing gigs came up so I've been a bit swamped.
Here's my spine & legs test. I used a couple different sensor configurations just to make sure they were behaving consistently.
As noted above, in Part 1 from 0:00 to 3:12 it seems like the bending is somewhat off axis. According to my compass, this should be working, but I'll do as you suggest in the future, and put down some tape on the floor to mark where the axes align. However, the legs looks to be bending nicely.
In Part 2 from 3:13 to 5:22, I reloaded the Blender file and tried again. This time I encountered more of the same (yet seemingly more pronounced) rotation problems with the spine, but also somehow the neck ended up twisting around and the left leg tended to swing wide.
I reloaded the Blender file again for Part 3 (5:24 to 7:34), no weird neck or leg problems this time (in fact the legs seem to move quite well), but again the spine was bending off axis.
And lastly, in Part 4 (7:35 to 10:14) I tried a this configuration using different K-Ceptors. The off axis spine issues seem to be a little less than with the first set of K-Ceptors, although the left leg bent a bit oddly.
Next up was the full biped test. Again, as noted above I tried this in a different room, but was getting the same sort of results as last time, except reloading the Blender file no longer fixed the erroneous rotations. And as you'll see the Notochord program was only using a small percentage of the Pi's CPU, and was only using a small amount of bandwidth on my router.
Questions and Some Speculation:
1 - How do I turn on matcap in the spine test Blender files? I've tried multiple things, but nothing seems to work. I figure that having matcap turned on might make it easier to spot odd rotations in the future.
2 - If my sensors were still exhibiting constant rotational drifts, would this interfere with the data from my movements and cause these incorrect and drifting rotations?
I ask because after I did these tests, I did some tests using the Test Cube file, and noticed that whenever the X and Y local axes of the K-Ceptor were being pointed along the global Z-axis in Blender, slow but steady drifts tended to occur.
However, I was about halfway through my K-Ceptors when I remembered that I shouldn't be testing the sensors a couple feet from my computer, so I'll have to redo those tests to confirm. Also, I think I've asked this before, but if I accidentally test K-Ceptors close to an EM source like a computer, do I have to recalibrate them afterward? Or will just retesting further away from the computer work? And if they are still exhibiting rotational drift when I test them properly, what do you think would be the issue causing this?
3 - I've been wondering about storage and calibration loss. For instance, I noticed that the Perception Neuron comes with an Anti-Mag case for its sensors, whereas I'm merely storing everything in a Fallout 76 canvas bag on my equipment shelf:
Would changing my storage method help with some of these issues?
Also, from what I've heard, the Perception Neuron holds it's calibration for about 30-minutes and then it needs to be reset. Does K-Ceptor calibration have a similar time limit?
4 - Is there some sort of debugging or verbose mode I could activate in the default_biped Blender file that would give you some additional data that might provide more insight into these issues?
5 - Ok, this is kind of an out there question, but what is the chance of implementing magnetic immunity? I was watching some Corridor Crew videos on YouTube and saw that they used an Xsens mocap suit. In a promo video on their website, they show this off with a person wearing their suit dancing around shipping containers. However, they said that discovering how to implement magnetic immunity was the result of 15 years of research, so I'm guessing this wouldn't be a possibility, but you know more about this sort of thing than I ever will, so I figured I'd ask.