The feature to include the gyro terms in the solver is excellent - it turns out to be really important for correct simulation of aircraft spins and snap rolls etc, so much appreciated. I'll post a video of it in my R/C flight sim shortly.

Thanks - Danny

Have you posted your video already? If yes, where can I view it?

Ooops I forgot! https://youtu.be/LwjFUzsymjE

The manoeuvres most affected are the flat spin right at the end of the video, and also the "flips", for example around 2:25 - if they don't start perfectly aligned with the wing span then the glider "wobbles" out after a few rotations (not shown here). Before I included the gyro term it would happily continue to rotate around a slightly "off" axis, which is wrong.

Here's the designer of the plane flying the real thing (including the gyro terms!): https://youtu.be/6VZ0tRrXnKA

The feature to include the gyro terms in the solver is excellent - it turns out to be really important for correct simulation of aircraft spins and snap rolls etc, so much appreciated. I'll post a video of it in my R/C flight sim shortly.

Thanks - Danny

Have you posted your video already? If yes, where can I view it?

The feature to include the gyro terms in the solver is excellent - it turns out to be really important for correct simulation of aircraft spins and snap rolls etc, so much appreciated. I'll post a video of it in my R/C flight sim shortly.

Thanks - Danny

So what's on the plate for the future? I see some work in featherstone btMultiBody stuff...

I also see some logic for tracking 'islands'.. which I guess are groups of related things that might collide.
I would think that as an extension my crazybulk review by thisfitlifestyle to that one might order collisions based on their collision with static bodies; such that if a body collides with a static body it gets an additional flag set in it that indicate it's a static collidable; would think that would solve issues stacking heavy things on light things. (been doing a lot of reading and now see where that artifacting comes from). Then the 'roots' of islands would process first and then tree out from there... but maybe then 2 bodies on a static ground next to each other would still have issues.

Also digging through collision dispatch and DISPATCH_CONTINUOUS is never actually used? Isn't that kinda what discretedynamicsworld/CCD would need? to precalculate the penetration depth to determine a more appropriate time step before doing the discrete resolutions?