Resolved RigidBody.AddForce vs PhysicsVelocity.Angular question

Hello,

I've been tearing my hair out for a long time trying to fix what should be a simple world -> local transform (as far as I know).

I am trying to write an aerodynamics simulation in Unity ECS. In the original code, everything works. It does the following:

Code (CSharp):

1. rb.AddForce(currentForceAndTorque.p);
2. rb.AddTorque(currentForceAndTorque.q);

AddForce and AddTorque both apply in world space, and the currentForceAndTorque values are calculated in world space, so it all works fine.

In my code, when I apply my lift and drag with

Code (CSharp):

1. physicsVelocity.ApplyLinearImpulse(physicsMass, forceAndTorque[0] * deltaTime);

It works fine. ApplyLinearImpulse is in world space, so it's all good.

Initially, I was trying to apply the angular velocity like this:

Code (CSharp):

1.  
2. physicsVelocity.Angular += forceAndTorque[1]  * physicsMass.InverseInertia * deltaTime;
3.  

This didn't work -- physicsVelocity.Angular is in local space but my forceAndTorque[1] is in world space. So I tried a lot of different variations of this:

Code (CSharp):

1.  
2. quaternion inverseWorldRotation = math.inverse(vehicleLtw.Rotation);
3. // Convert world-space torque to local space
4. float3 localTorque = math.mul(inverseWorldRotation, forceAndTorque[1]);
5. physicsVelocity.Angular += localTorque  * physicsMass.InverseInertia * deltaTime;
6.  

All of them had my plane go totally haywire at the slightest whiff of (non-x-axis) rotation.

Now I am trying this:

Code (CSharp):

1.  
2. var currentWorldAngularVelocity = math.mul(physicsMass.Transform.rot, physicsVelocity.Angular);
3. physicsVelocity.SetAngularVelocityWorldSpace(physicsMass, physicsMass.InertiaOrientation, currentWorldAngularVelocity + worldTorque * physicsMass.InverseInertia * deltaTime);

[/code]

With pretty much the same result.

I've also tried pretty much every other possible quaternion available in place of `physicsMass.Transform.rot`, e.g. physicsMass.InertiaOrientation, and the LocalToWorld on the entity as well. I've completely run out of ideas of how to approximate "rigidbody.AddTorque" in ECS.

I'd really, really appreciate any pointers as to where I'm going wrong here. What I'm trying to do seems so simple, just "AddTorque" but for ECS, but I cannot find anything that is equivalent.

Update:

I've now also tried

Code (CSharp):

1.  
2. physicsVelocity.ApplyAngularImpulse(physicsMass, forceAndTorque[1] * deltaTime);
3.  

and this also seems to apply the impulse to the wrong axis, also causing the plane to rotate wildly.

Thanks

 

@FareedDudhia

I am just curious.
Did you run by any chance into NaN infinite torque and force, in your aerodynamics code conversion?

 

@daniel-holz

Thanks so much for this answer, after much digging I found and solved the issue using some of the code in the examples you screenshotted above. My only remaining question would be how can I browse this code myself? If I could have found the code you linked myself I might have been able to figure it out on my own.

@Antypodish
I did, yeah, but my issue in the end was simply that a value for angular velocity was in local/motion? space when it should have been in world space. I can clean up the working code and link it to you if you'd like (might take me a while).

 

Oh that would be amazing. I looked into same repo, or one of its fork. I have converted it, but had mentioned physics incompatibilities.

No worry, I will appreciate for the repo, but I am not in a hurry, so do whenever you will get he time.

Many thx in advance.

 

Hey,
Just a little poke, in case you are busy, or forgot
No rush however.