🎉 Celebrating 25 Years of GameDev.net! 🎉

Not many can claim 25 years on the Internet! Join us in celebrating this milestone. Learn more about our history, and thank you for being a part of our community!

Back to Math and Physics

Object vs ground constraint solver

CasualKyle · 2021-12-11T22:38:25

I'm trying to implement an impulse based constraint solver for the case of a 3D cube falling onto a horizontal plane. When a vertex of the box is below the plane, I'm solving the constraint where the vertex of the box should be anchored to the contact point on the plane i.e. a ball and socket joint. In this example I'm solving the constraint where b - c = 0. Since my constraint solver works at the velocity level, the idea here is that I'm applying an impulse to the cube at point b. This will both move the cube upwards and induce a rotation. Also, there is no restitution involved. I'm trying to gain an intuitive understanding about whether my solver is working correctly or not. If I set the baumgarte stabilization coefficient to 1 such that the constraint is completely satisfied in one run of the solver then the cube should move & rotate such that b - c = 0 right? The problem I'm having is that this is not happening. The point where the red line intersects the plane would be point c in the first diagram and the corner of the cube would be point b. This is the result of solving the constraint once with a baumgarte stabilization coefficient of 1, in that case is this correct? I don't think it is since b is not equal to c but I've checked the math and I cannot find a mistake. Because the math is so complicated, before I share any of the code which calculates the impulse is the assumption correct that these two points should be equal? Should the direction of the impulse be completely vertical? Or is it expected that it would have some non zero horizontal values to account for the rotation applied to the box? Here is a breakdown of the code. // integrate linear and angular velocity box.velocity += ...; box.angular_velocity += ...; // integrate position and rotation box.position += box.velocity * delta_time; let w = Quaternion::new(box.angular_velocity.x, box.angular_velocity.y, box.angular_velocity.z, 0.0); box.orientation += w * box.orientation * 0.5 * delta_time; box.orientation.normalize(); if box vertex is below plane { // calculate force required to solve the constraint let f_linear = ...; let f_angular = ...; // integrate linear and angular velocity box.velocity += (1.0 / mass) * f_linear; box.angular_velocity += inertia_tensor_inverse * f_angular; // integrate position and rotation box.position += box.velocity * delta_time; let w = Quaternion::new(box.angular_velocity.x, box.angular_velocity.y, box.angular_velocity.z, 0.0); box.orientation += w * box.orientation * 0.5 * delta_time; box.orientation.normalize(); }

Math and Physics Programming constrained-motion

Started by CasualKyle December 06, 2021 11:31 PM

10 comments, last by CasualKyle 2 years, 6 months ago

CasualKyle

Author

December 11, 2021 10:38 PM

Okay I think I understand what your saying. The dV' I solve for is the change in velocity of the contact point which is linear. It's an instantaneous velocity. But, when I apply the forces to the body that contact point doesn't have a linear trajectory since I'm applying a torque.

muleax said:
This effect should be less noticeable when term r x w is small: smaller penetrations (usual case) or, for example, if you rotate the box on 45 degrees and collide with horizontal ground.

Yep, I've noticed this is to be true. It makes sense, if the box was rotated 45 degrees then the contact point would have a linear trajectory.

I appreciate all the help you've given, thank you!

🎉 Celebrating 25 Years of GameDev.net! 🎉

Object vs ground constraint solver

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

🎉 Celebrating 25 Years of GameDev.net! 🎉

Object vs ground constraint solver

This topic is closed to new replies.

Popular Topics

Recommended Tutorials

Reticulating splines