NeoTokyo_Nori wrote:Happy new year.
@konstantin_lozev I checked the video, and it looks like you managed to get the drift under control, which is great. Your glovePIE script looks long, and I havent used glovePIE for a while, so I am not sure if I will be able to understand it. But it would be great to be able to port this over to FreePIE.
cheers.
Hi, Neo, Happy New Year to you too
. Yeah, I am pretty sure the pitch drift is safely under control within the cage of the accelerometer data, if I can use the metaphor, which explains quite well what it does actually
The part of the video that is with the rectangle (cube looked from the side) may seem choppy, but it only affects the rendering of the cube and is due to the fact that Crysis 2 runs in the background and not because the alignment makes the mouse movement choppy (which is evident from the in-game part after that). You are right, the script is long, but it's mostly window dressing, the core is quite simple and small actually. I cut the most relevant lines below and simplified them a bit, if you want to have a quick idea:
---------------------------------------
var.delta = (Removeunits(wiimote.PitchSpeed in radians per second)*Removeunits(Delta(timestamp)))// this is the variable that stores the changes to be made to the pitch at the end of the frame, it absorbs the effect of the gyro readinds here and later gets corrected if the alignment of gyro and accelerometer data is triggered for the particular frame
...
if abs(wiimote.gz)>1 { // this normalizes gz to 1
var.corrgz = sign(wiimote.gz)*1
} else {
var.corrgz = wiimote.gz
}
if abs(wiimote.gy)>1 { // this normalizes gy to 1
var.corrgy = sign(wiimote.gy)*1
} else {
var.corrgy = wiimote.gy
}
...
if (abs(var.corrgz)<=var.sincosswitch) { // this switches from aSIN to aCOS when > +-45 degrees. It appears that aSIN,aCOS gives least noise around +/- 90 degrees,
if (var.corrgy>0) { //this is the check that is meant to enable the correction of angles beyond 90 degreess
var.cumulpitchspeedacc = aSin(var.corrgz) in radians // var.cumulpitchspeedacc = aSin(var.corrgz) in radians is the acceelerometer-measured pitch for pitch below +/-90
} else {
var.cumulpitchspeedacc = sign(var.corrgz)*pi()-aSin(var.corrgz) in radians // this is the correction for andles beyond 90 degrees
}
} else {
var.cumulpitchspeedacc = sign(var.corrgz)*aCos(var.corrgy) in radians // sign(var.corrgz)*aCos(var.corrgy) in radians is used for angles > +/- 45 degrees
}
...
var.deltaaccgyro = var.cumulpitchspeedacc-var.corrcumulpitchspeedgyro // this measures at each frame the difference between the accelerometer and gyro readings
var.deltacumulpitchspeedgyro = var.oldcumulpitchspeedgyro - var.cumulpitchspeedgyro // this measures the deviation of the gyro measured pitch since the last point where it was aligned to the accelerometer data
...
if (abs(var.deltacumulpitchspeedgyro)>var.alignmentmarker) { // this sets the trigger for the alignment section of the script
var.startcorr = true
}
...
if (var.startcorr = true) { // this activates the script that gradually alignes gyro to accelerometer data when the gyro readings have moved beyond the marker since the last alignment.
if (Sin(abs(var.deltaaccgyro)) > var.deltatolerance) { // this stops the alignment after the difference between the accelerometer and gyro readings is decreased under var.deltatolerance
var.corrstep = abs(var.deltaaccgyro/var.corrstepnumber) // this sets dynamically each frame the extent of the correction/alignment towards the accelerometer data. The bigger the difference between gyro and accelerometer readings var.deltaaccgyro, the bigger the alignment step per frame (which means the alignment is actually decelerating over time, which is what we want). With the var.corrstepnumber you can tweak yourself that deceleration further. The default is 5 (float, so can also be e.g. 4.7), if you want faster alignment, set it below 5, of you want smooter, but laggier alignment, set it above 5.
var.delta += sign(var.deltaaccgyro)*var.corrstep
} else {
var.oldcumulpitchspeedgyro = var.cumulpitchspeedgyro // this resets the dybamic marker after the algnment has finished
var.startcorr = false // this switches off the aliggnment untill the next marker
}
}
...
var.corrcumulpitchspeedgyro += var.delta // this corrects/aligns the gyro readings to the accelerometer reading by assigning the corrected delta to the final pitch variable
...
if (odd(var.switcher) { // The check enables the switching of the wiiimote tracking
var.fake -= Removeunits(var.delta)*var.sensitivity // var.fake is a value that retains the real data that is transferred to mousedurectinput in order to assign it to mousedirectinpit each frame, thus avoiding accumulation of rounding
Mouse.DirectinputY = var.fake //this assigns the corrected updated non-rounded pitch (var.fake) to the mouse. The MousedirectInput accepts only whole numbers, so it rounds out this assignment. Therefore, the use of the var.fake is necessary
} else {
var.fake = Mouse.DirectinputY // resets the difference when the script is switched off with the home button (changes var.switcher)
}
---------------------------------------------
As you can see, there are 5 components to it:
1. Reading and storing the gyro data into var. delta
2. Converting the accelerometer data into pitch using aSIN and aCOS to obtain var.cumulpitchspeedacc
3. Triggering alignment to accelerometer data after the gyro has detected change in pitch beyond 0.05 radians (abs(var.deltacumulpitchspeedgyro)>var.alignmentmarker)
4. Gradual decelerating correction of gyro-measures pitch towards accelerometer-measured pitch (var.delta += sign(var.deltaaccgyro)*var.corrstep)
5. Accumulation of var.delta to a fake mouse variable var.fake and then assignment to the Mouse.DirectinputY. This additional step with the fake mouse variable is necessary because Mouse.DirectinputY accepts only whole numbers (mickeys), so it disregards the remainder at each frame, and that remainder accumulates over time leading to unnecessary drift, which is completely avoided by accumulating var.delta in the fake mouse variable. I think the example FreePIE mouse script may suffer from the same issue (although I have not tested it), which could be solved the same way.
Now, on my plans for conversion to FreePIE - my major hurdle is Python since I am only an enthusiast, so I never actually learned any Python (GlovePIE was more accessible to me since I gobbled a lot into Adobe Atmosphere, which used JavaScript, as well as some DHTML 15 years ago). So the plan is I learn first Python... I wish there were Google Translate for code
