1 reply to this topic

[irlanrobson]

I'm implementing a physics simulation for my final course project and updating the rigid bodies like this:

 

calc_forces(); //calc constant forces and torques

 

velocity += (force / mass) * dt;

position += velocity * dt;

 

ang_momentum += torque * dt;

orientation += ( (ang_vel * orientation) * 0.5f ) * dt;

 

orientation.normalize();

 

ang_vel = inverseInertia * ang_momentum;

 

 

//IF COLLIDE WITH OTHER

 

velocity += ( (j * collision.cn) + (u * j) * collision.ct ) / mass;

ang_vel += pt1.cross( (j * collision.cn) + (u * j) * collision.ct ) * inverseInertia;

 

other.velocity += ( (-j * collision.cn) + (u * j) * collision.ct ) / other.mass;

other.ang_vel += pt2.cross( (-j * collision.cn) + (u * j) * collision.ct ) * other.inverseInertia;

 

I've read the Baraff's papers but i think that what i'm doing is "no change in angular momentum" and in this moment I don't have any idea how to calculate.

Obs: when i change ang_vel to ang_momentum, the results are better but i want correctly results. 

[EWClay]

The only obvious problem is that you are changing angular velocity in the collisions and overwriting it in the update, so the change would have no effect. As you seem to realise, changing angular momentum is better.

If it still looks wrong, there could be plenty of reasons that are unconnected to the code above.