DOTS Physics: Dynamic objects will not settle on other dynamic objects

As demonstrated in the gif below, any dynamic physics body that comes in contact with a static physics body, stop as intended with the expectation from Neuton's 2nd law; however, when a dynamic body attempts to settle on another dynamic body, it seems to fight back a bit resulting in bouncing effects. This happens with varying values of friction and restitution as well. It occurs in new empty scenes with minimal components as well as the samples provided.



Question: Is this a bug? Or is there something we need to do in addition to handle multi dynamic object collisions to create, for example, piles of objects?

 

An example of the physics sample of this occurring:

 

Can you validate that the physics sim really runs on fixedupdate and not regular update? Many ECS systems are meant to run by default in fixedUpdate world in the future versions of Unity, but I think right now this whole simulation/presentation worlds thing is not ready yet and they might just be running on Update

 

Yep, it does run in Update for now

 

Yes they in Update and this why you can see in examples, sometimes simulation faster or slower, depend on framerate It looks funny

 

https://blogs.unity3d.com/2019/03/1...eneralpromo-2019-03-19&utm_content=blog_havok

The blog post indicates that the Havok solution coming later this year will have support for stacks of dynamic objects. So in Unity physics this probably wouldn't be considered a bug. More like lack of a feature. If it's something you can implement or work around yourself, the good news is that it's an open source package.

 

That's a good sign, because it means the bug is easy to solve! (Maybe)

Simply making physics run on fixedUpdate might slove stability issues. This is most likely something unity is aware of and they were just waiting for the simulation/presentation world setup to be ready

Since UnityPhysics wants to be deterministic, it is literally impossible that it was actually meant to run on update

See this post: https://forum.unity.com/threads/why-is-simulation-the-default-system-group.639058/#post-4298449
right now we are in the "next" release of Entities that he speaks of, where FixedUpdate doesn't work yet

 

I wouldn't expect physics moving to a fixed update loop (whenever that even happens) to fix this. Referring again to the blog post, it implies that the lack of caching in the Unity physics package limits suitability for "complex scenarios" such as stacks of dynamic rigidbodies.

 

It will. All simulation group will be moved to fixed update (it was actually but changed in 0.0.28 ECS preview but temporary) https://forum.unity.com/threads/why-is-simulation-the-default-system-group.639058/#post-4289911