Interpreting ContactNormalB value returned by getContactPoints

[shivanimall]

I tried multiple objects to simulate collision, and also created angular and linear velocities along all axes using resetBaseVelocity. However, when I try to get the contact collision forces, the third value(Z direction) is only populated(with -1) in the vector returned by ContactNormalB as following: (-0, -0, -1) from getContactPoints. For the first 2 values, I am always getting 0.

I could see frictional forces along X/Y directions. But, how should I interpret the value of ContactNormalB? From my understanding if an object falls at an angle, there should be some force values along all axes(X, Y, Z). So are there no contact normal forces along X/Y directions? which seems incorrect.

Please let me know if my question is unclear. Thanks

[shivanimall]

Hi @erwincoumans following up on this, in case my question is unclear, please let me know.

[xissburg]

The contact normal B should be a vector in B's object space. To obtain it in world space you have to apply B's rotation to it. I know this on a conceptual level, not sure what the code would look like in your situation but I think you can figure it out with this information.

[erwincoumans]

ContactNormalB is in world space, pointing from B towards A. If an object falls at an angle at a flat surface, the contact normal can still be [0,0,-1], since the contact is generated from a vertex-face contact. There are no normal forces along X/Y, only friction forces along X/Y.

[shivanimall]

thanks for the response @erwincoumans
noted the point on vertex-face contact. is it possible to point me to a demo/example that can generate normal forces along X/Y directions as well? any simple free fall object-collision simulation on the plane would be great.

hoping to see simulation cases that I am not capturing in my examples so that I calibrate my use case. thanks again.