It seems like the initial rotation of the meshes is not being taken into account when building their bounding boxes. It might be an issue with the physics engine, but might be an issue with the export itself. What happens if you disable the physics impostors? are the bounding boxes the same? Have you tried using a different physics engine?
And as @sebavan said, reproduction on the playground would be helpful for us to understand the issue.
Thanks for looking into this for me, please find the link to the playground as requested below.
All looks well in the playground except for the collider for the small mesh(railings3). Now I’m thinking it might be the version of Babylon I’m currently running or perhaps something else.
skipLibCheck fixed the errors and the rotations issues are now resolved but the position is off for the colliders. The playground code already shows the position issue. see code below.
Seems to be an issue with the oimo impostor initialization. Oimo has not updated in a while, and while we do support it, it tends to be the lease used engine in the last few years. Have you considered using a different engine? Cannon seems to calculate it correctly, and so does ammo:
Thank you for the assistance guys, I will move to ammojs for now and hopefully all works and if something unusual pops up, I will let you know. Also, CannonJS example doesn’t look right, is that the expected result @RaananW