Flypt Filter Research (or why doesnt my rig move right?)

Let me describe the concept in summary.
We get from game telemetry acceleration (a), data that represent the human sensation.F~a
If we want to map to a position, we have to integrate 2 times over the period of travel, that is the behavior of real world.
In lack of travel and time we take a HP to LP combination.
The HP reacts as differentiator, the LP as an integrator.
So first we get da/dt or a` that represents the variation of acceleration.
The key point, if we do the reverse, integration, we integrate da/dt and lose the constant a(0) (constant acceleration)
a number the integration can not deliver. The same for further integration V(0) and S(0) but this values are not relevant creating no forces.

Why do we not use the speed in contrast acc. ?
See example diagram of longitudinal speed with HP/LP. (grew acc, red speed with HP/LP Filter)

Can not agree, imagine a race car very hard braking, the driver hanging in seat belts, decrease or loses the brake a bit , still remains hanging in seat belts, no acceleration from back.

that is correct method , as long as no false cues are created.

To a limited extent false cues are unavoidable and OK. Don't get me wrong here, it makes total sense to mitigate those as good as you can, but given the constraints (space, money, etc...) you will not be able to create a perfect match.
And you don't need to. As @noorbeast put it nicely, all we need to do is...

Humans are not precise physical measuring instruments, capable of processing measurement data and deriving exact values from it. Human perception takes plenty of shortcuts to reach "reasonable" conclusions about our environment. Just barely enough to allow us to procreate.

Whenever I hear: "This is obviously wrong" then the reason for it being so "obvious" is that the occupant of the rig has not properly isolated himself from outside stimuli. It wouldn't be so obvious if outside vision, hearing and sense of tactile feedback were properly taken away from him, as it should on a motion rig.
Also, you have only ever been considering surge, or any other translational DOF. But when combined with +0.5G from tilt it may very well be that the surge DOF moves back at -0.2 G while the driver still feels positive 0.3Gs of acceleration. Tilt and surge play together!

For the case of the braking race car driver: If you brake at say 2G, your head will tilt forward slightly, your arms and shoulders will move forward a little and your chest will feel the pressure of the harness straps. But after just a second your muscles will intuitively adapt to the new normal: Your neck muscles will pull your head back into position, your arms and shoulders will be repositioned to give you good control of the car and your breathing muscles will adapt to cater for the situation your body finds itself in now.
Now, if you suddenly reduce to 1G braking, you will slam your head back against the headrest, your shoulders will move back a bit too far and you will breath in freely like there's no pressure on your chest at all.

The change of a state is perceived much stronger than the state itself! And if you isolate yourself from outside stimuli and put yourself under some stress your self perception will be tricked. It will not be perfect, but it's good enough to provide you with meaningfull training.

Cheers....

Except Niki Lauda! Niki Lauda!

I found a nice example. Go to 1:02.
-->Link<--

Yes, this is a Hollywood movie, but afaik they actually shot Tom Cruise off a carrier for this scene. During the cat shot the aircraft accelerates at 2-3G. When the aircraft leaves the deck it will only accelerate at ~1G from the afterburners. See how his head slams forward when the acceleration drops from 2 to 1G?

I know it would be better if we could recreate every force to the precise value, but within the constraints it is an OK compromise, I'd say.

i would interpret this scene as that the short negative da/dt is so extreme that this pulse appears as a kind of shock wave.

want to add some further thoughts.
If a force is propertional to acceleration (F=m*a) and two forces interacts to a body in opposite direction, a movement, or a direction change of movement could only be happens if one force is greater than the other.So the same should apply to acceleration.
Just done a test drive record and view long. Acc. with genius flypt, by doing a manual gearing up and some intermittent heavy braking.
The intermittent braking shows no acceleration.(curve negative)
Gearing up however shows deceleration peaks, this because the clutch cuts the motor from drive so force and acc. is zero and if it engaged the load of wind, drive mass decelerate in a short moment.
The same could be done with a flight sim to check the behavior in the tom cruise video.

You can do a cat shot off the carrier in DCS and observe the longitudinal acceleration in Mover. I'm sure you will see the acceleration drop significantly, but not go negative, as long as you have significant thrust,... which you should

Just created a method like in the sketch for testing the filter performance.
Viewing the filtered output of a pose with a downstream double derivative, EMAHP(EMAHP(value:100);100) filter.
A Triangle filtered with that combination delivers a plausible result. Basic curve triangle, first derivative square, second the blue shape.
Both curves, compared in graphic viewer regard to different signs in shape, show the amount and size of false cues.

That looks plausible to me

It highlights something really interesting:
If you increase the frequency of the triangle-wave you should see that the gray area (output) and the triangle-wave (input) become more and more similar. That's because the HP filters allow more of those high frequencies to pass. In the edge case of an insanely high frequency the peaks and throughs should almost align.
Can you confirm that?

Every periodical signal, as the triangle, is in the frequency domain a sum of sinus waves. A square have less coefficients as the triangle. If the Filter removes that elements a triangle becomes a square.
If a HP-Filter passing to much of low sinus elements e.g. HP(10000HP10000)) the output approach the input.
My aim of the second, the triangle picture should demonstrate that the HP(100)HP100)) seems to be a proper choice for 2 time derivative, as a greater filter value of HP(500HP500)) makes the edges of the square the first derivative already to round, a HP10(HP10)) removes to much in amplitude.
I would suggest to compare both curves with real sim data.

@Dirty if you're ever up to it I'd be eagerly awaiting video #13 in the series. Acceleration and translation DOF I get. But rotational DOF motion cueing using Roll/Pitch speed still escapes me. I cannot get any satisfying result from it :and it completely escapes my mind.

PS: thank you for the first 12 videos, they were very enlightening!

As a rig works as a position system, basically without processing only values with the same dimension fits mathematical to each other e.g. meter to meter or degree to degree. Not taken native degree to rotational axis e.g. roll is justified of missing centripetal forces in static rig in contrast to dynamic in real where centripetal and g-force are geometrical vector added.
The fact of dynamic to static movement also apply to other axis, e.g. yaw.
The use of speed followed by HP is practical the same as acceleration. (a=dv/dt) that presents the base for the following TP to create a motion cue not presenting the real motion or forces itself.
Imo. translation uses da/dt as a fact of constant acc. over a long period, however rotational can use dv/dt as the rotation is normally limited.

For my use case I figured it out I think. As I currently have a pegleg platform with 3 actuators - albeit being relatively close to COR (my butt is on the floor and pretty much at COR and my head just ~550mm above and ~150mm behind it) anything that's high frequency just feels like false sway/surge cues and like sitting on a rocking chair.

I have now set a ButterworthLP with a cutoff of 0.5Hz for the roll and pitch speeds and it feels much better than with the 5-20Hz I typically see in hexapod profiles. It basically feels like pitch/roll angle but without a lot of the false cues they provide and relative to the other movements vs. the absolute nature of pitch/roll angle.

Sorry to interfere, “centripetal in real”… is the centripetal force from the perspective of cockpit not a simple linear vector? We thought about it a bit and the offset from the rotation point calculates in Mover a vector which you can translate to the rig in many scenarios. Ex. Yawing on ground in a A330 wil produce a significant SURGE force pushing you not only left right but also out of the seat. Following the equivalence principle the rig tilts also forward down with a simple but fancy math. I am quite convinced it’s a nice way how you could simulate centripetal forces.