So we have two bodies, a "plane" and a "ball". Is there a way to detect collision between them when the ball passes through the plane by having physics enabled on the ball? I know how to do it by when disabling dynamics or setting kinetics to 1, but that disable the physics of the ball. I need gravity.

Are there any cunning @filter settings? Or maybe some trick with two jit.phys.worlds where one would serve only for gravity and position matrix while the other would have dynamics off, react to position matrix from world one and detect collisions?

collision-data-without-actual-collision-while-physicsdynamics-enabled

But there is one more thing that would make me about 50% happier person: is there a chance to "turn on" the impulse? The impulse is constantly 0? Basically I would like to know with what velocity one body "hits" (*passes) through the other...or what force would it be if the bodies would actually collide? I just need some kind of intensity of ...

I am returning a slightly modified patch:

I guess I could get what I need from the duration attribute. But since ideally my object of collision would be only a point or a surface i need some very thin cube. Now if I set the scale parameter of the cube to 0.02, which is the smallest size in order the collision to be detected (with the default parameters in the patch), then the duration is always 0. With the size of 0.07 I already get some duration but that doesn't work for some faster passings...(*if you increase the height/starting point of the falling object the detection does not work any more). Are there any optimal settings for what I want to achieve or maybe some better ways to do it?

There is no impulse because the ghost has no mass. It is "on."

Very small dimensions could result in extrememly short or missed collisions (between calculation steps of the physics engine).

The physics body of the ball can report its velocity. You could get the velocity at the moment of the collision.

There is no impulse because the ghost has no mass. It is "on."
Very small dimensions could result in extrememly short or missed collisions (between calculation steps of the physics engine).

The physics body of the ball can report its velocity. You could get the velocity at the moment of the collision.


Thank you SLO for your help! That is exactly what I needed!

But I would anyway like to have a better insight into the physics engine. I thought that the case why some events are missed (smaller dimensions and higher velocities) lies in the fact that the collision might happen between two frames so there is actually no collision at the sampling intervals determined by the frame rate. So I slowed the frame rate down to 1 fps and I was surprised that in a case when collision is not detected the two bodies are actually colliding/overlapping at the time of "frame event".

So can anyone explain me why the physics engine is not detecting a collision although the overlap of two bodies happen exactly at some specific frame? Does two bodies need to overlap for two frames? Is physics engine maybe totally separate from the frame rate?

I would really need to to know this since I need to squeeze as much accuracy as possible from these collisions - because I need to work with small and fast objects.

You should look into the "fixedtimestep" and "maxsubstep" attrubutes. I believe there are some useful forum posts on these.

Collision data without actual collision while physics/dynamics enabled