PhysicsDirectBodyState2D: Reported values of `get_contact_impulse` not as expected

stephanbogner commented 7 months ago

Tested versions

Tested in:

v4.2.1.stable.official [b09f793f5]
v4.0.4.stable.official [fc0b241c9]

System information

Godot v4.2.1.stable - macOS 14.1.2 - Vulkan (Forward+) - integrated Apple M2 Pro - Apple M2 Pro (10 Threads)

Issue description

Context:

I try to make breakable objects and wanted to use get_contact_impulse for this case do detect "hits".

Problem:

How strong the impulses are seem almost random ... but is hard to test in a reproducible way

My physics lab setup 😂:

All boxes are the same (mass, gravity scale, etc.)
The top text shows the highest recorded accumulated impulse (sum of all impulses in one _integrate_forces-cycle)
The lower text shows the highest recorded speed (length of linear-velocity)
Distances setup:
- The first box has no distance to the ground
- The second has 1px distance
- The next three are arbitrary
- The last two have the same distance to the ground

Outcome:

Velocity does not linearly correlate to impulse (but it should ... I think?):
- Box 1 has no speed but an impulse
- Box 2 has a impulse/speed ratio of 1.55
- Box 3 has a impulse/speed ratio of 1.19
- Box 6 and 7 don't have the same impulse but I guess they should as they have the same fall height and weight (since I sum impulses)

Note

I am not 100% sure if these are bugs or I am misunderstanding the concept of what get_contact_impulse is reporting.

Steps to reproduce

I recommend making a similar physics setup and playing around. Make sure you turn on contact_monitoring and set a value for max_contacts_reported.

I used the following code to track values:

extends RigidBody2D

var highest_speed := 0.0
var highest_impulse := 0.0

func _physics_process(delta):
    var speed = linear_velocity.length()
    if speed > highest_speed:
        highest_speed = speed
        $Velocity.text = str(snapped(highest_speed, 0.1))

func _integrate_forces(state):
    var total_impulse := 0.0
    for i in range(state.get_contact_count()):
        var collided_obj = state.get_contact_collider_object(i)
        var impulse = state.get_contact_impulse(i).length()
        total_impulse += impulse

    if total_impulse > highest_impulse:
        highest_impulse = total_impulse
        $Impulse.text = str(snapped(highest_impulse, 0.1))

Minimal reproduction project (MRP)

bug-contact-impulse.zip

stephanbogner commented 7 months ago

A different setup (bug-contact-impulse.zip):

All objects have the exact same wonky shape, weight and mass ...
... but are rotated a bit (between 0° and 5°)...
... which makes the impulses differ widely on impact

rburing commented 7 months ago

Thanks for the MRP!

I haven't looked into this myself but others have made PRs for it. Can you test the PR https://github.com/godotengine/godot/pull/81654 and comment on whether it fixes the issue?

stephanbogner commented 6 months ago

@rburing Thanks for the reply.

I built from source (the other approaches mentioned on the page didn't work) and now generally it's much more in line with what I'd expect as an outcome ... but since I don't know enough about actual physics it's hard to say whether it's correct now or not though, but below are the results from my tests.

If you have a more clever idea for how to test properly I'd be happy to make more tests.

godotengine / godot