FlyPT Mover

Source within a source would be great if I imagine correctly what you mean. It could look at any outlier data that it was suddenly hit with such as abnormal dive gLoads, and actively scale it down rather than clip it.

Yes it’s easy. For ground/air separation you can find your own solution or use my approach:

Weight on wheels minus current weight. That should make it aircraft independent.

Being limited with 3 extras It was a simple solution to leave the flight model unaltered and add a very well working rumble on heave axis.

As you don’t have any limitations, you could find a nicer (more complex) solution which maybe could include more realistic suspension behavior. (Roll, Pitch)

I am looking forward

@pmvcda

Here is a short clip of the 4 actuators rig reaction to test loop to illustrate what i said earlier.

Jerky movements and rig jumps to max displacement of test loop as soon as you start the loop. You should do something for test loop: slow start, or start loop at middle not at full displacement.

Right now if you rig is not set up correctly and you don't have safety measures like on Thanos controler, you can blow your actuator

Hi Flypt ;-) ,

Regarding the extra effects in plugin for car games , the traction loss is very useful . As yaw is world oriented and for me not useable I prefer use traction loss in yaw axis .
In the plugin you have in your software I can find only the yaw ( world oriented) . That means that traction comes from a calculation ( wheels ? ) or could you get also traction loss direct from game data ?
Thanks

I strongly believe is some bug in the serial code part, because the serial output on simtools is way more precise and smooth and when you define the 2ms delay its actually 2ms delay, not 40ms as it seems in your video!!!!

See this video where simtools handles the serial communications but still uses a FlyPT plugin for 6DOF math... Its absolutely smooth and accurate and fast. You can see the vibration from the road surface detail!!!!



In comparison when the motion is transmitted via the FlyPT serial output, the "Stepping" is clearly visible when the target values are far apart. The Servomotor is incredible fast and car reach the targets FASTER than it receives new positions and you see this stepping motion...
In the following video I actually set the delay to 1ms for the serial, which normally it might cause "skip data-sets" in Simtools where timing of the serial is faster.



This stepping is not visible at all on other systems that use DC motors or AC motors with PID and closed loop with sensors where the reaction speeds are slower and filtered by the PID loop itself.


Thanks
Thanos

I gave up for the moment has I can't do nothing with the serial module.
Too much risk to break something right now

I just hope he could sort something in the future.
@pmvcda If you want to further test or develop your software around 4 actuators rig setup you can join @Thanos Discord we will be happy to help and give Feedback

No problem. I might have it solved tomorrow. I'm at work right now, only later this night I will look at the code.

If physics are done correctly, the sum of all forces, (not only yaw), should generate the effect of traction loss.
So it shouldn't be needed for Hexapod rigs, where we have all DOF.
But even so, it's interesting to have it.
So I can add that effect as an extra pose (or maybe other solution).
TL would be a relation between lateral acceleration and longitudinal acceleration. The amount of rotation depends on that relation.
High lateral acceleration and low longitudinal acceleration would generate a rotation (well, in reality it could be just a lateral slide).
I have to think more about it.
For now, the basics have to work. Attack on serial output!

Thanks taking time to explain this to me ;-)
Haha your to do list is being to increase a lot ... understand that this is not the priority one ;-)

To help with the serial debugging, here a few things to consider:

1. Do you output serial data constantly in regular intervals? Like 2ms intervals if set to 2ms?
2. Does Serial has priority to execute the transmission intervals over other tasks?
3. Does it wait for motion cues to be calculated before sends new data to the serial port (asynchronous transmission)?

An indication that the serial code is not efficient is that it gets a bit smoother when you hide all the windows. If it has priority over other tasks, It should not affect its performance.

Also it seems the stepping is worse if the difference between two target positions in larger, indicating that it skips in between positions, thus the 2ms timing for example is not 2ms but much larger!!!


Thanks
Thanos

@pmvcda

I verified the issue with the transmission intervals on my controller in hardware using oscilloscope to measure the frequency of the complete packets...

-Simtools, 500hz frequency for 2ms intervals
-FlyPT, 64hz frequency for any interval setting up to 20ms... (then it drops more frequency)

See this video:




Thanks
Thanos

Wow, it's bad!
I'm going to take a look now. Must have done something stupid...

Of course we all agree “if physics are done correctly” everything should be properly simulated on a 6-dof.

Unfortunately, from what I wad told, many (over 80 percent) of racing games doesn't provide angular velocities.

On the top of it, SimTools „Standard“ puts the Euler angles on the main rotational dofs.

While the Euler can be usable as velocities „replacement“ for Roll and Pitch, the Euler on Yaw might be problematic because of the jump in compass value 360 deg. - 0 deg.

It could be, that a fix to overcome this “compass jump“ could be found so a yaw with a simple High Pass filter (washout) could do the job.

In order to demystify the TL which is usually present in the extras it would be good to provide some detailed info about how it is constructed.

For ex:
Traction Loss slider in simtools filtering remains a small mystery. Is it a special math to washout the data coming from the extra TL effect ?

@yobuddy , you are certainly the one able to answer and clarify all this question.

Thank you !

Edit:
Maybe a simple conversion 0/360 to 0-180 positive / 0 180 negative could do the job...

This may be a stupid question but are you using multi-core processing? I know Simtools does in order to achieve the speeds it does. Or it could be the correct term is multithreading.

@pmvcda I will be happy to help with making a tutorial video etc. once we get to where the software isn't likely to see big changes any more. Time is always a problem for me, but I will make sure to help as much as I can. One of the few things in life that I'm actually good at is explaining things in a way that is easy to understand. =)

@pmvcda
I've just been doing some X-Plane testing and noticed a couple of things.

1) In Simtools turning the Extra 1/2/3 must have a value greater than 0 entered so I used 1. I saw you had recommend to use 0 in there, but nothing will pass from Simtools.

2) <ignore this, I worked it out> On your pose: motion screen: See Sway and surge in the attached image, you can see they match. You probably have Longitudinal Acc mapped to the Lateral Value menu item as they move at the same time.

3) Extra 3 in the SimTools tuning centre for X-Plane should map to Yaw Acceleration out of simtools. I need this for Yaw cues. Nothing happens here even when I drop the input gain. How do we manually map Tuning centre outputs of Simtools to Pose motion inputs? Is this done in sources? (so I used the yaw source in Simtools and put it to use Extra 3 and it didnt seem top work). Anyway, it would be good to get all 9 values from Simtools.

4) To get the data from simtools tuning centre I set the Extra 1/2/3 to 1, I notice that the output values are usually way less than 1 (eg 0.02 or less). By increasing the gain in the Pose/Motion window to compensate, are we reducing the resolution of the original input signal to make a low rez output? <actually I just noticed Extra 1/2/3 are not being picked up from Simtools - I think that's what I need to learn>

Little update for those waiting for me to test things @pmvcda @Dirty @D.C. Moore

I began my night with a "quick" test of leap motion VR controller emulation. The "quick" test I planned took six hours, was an abysmal failure, and made such a mess of things that I reverted windows back to an image I made several weeks ago.

So, unfortunately that is all I got done tonight. Never even got to fire up the motion sim. I'll try again tomorrow, but for now it's just about bed time. =(

@Trip Rodriguez
Are you trying Leap motion so you can use the switches in cockpits etc with your natural hands and no controllers?