m_appliedImpulse is 0 for a Generic6DofConstraint

[ramsampath]

Hello,

The m_appliedImpulse member seems not to be updated in the case of a Generic6Dof constraint but seems to be updated correctly in the case of all other HingeConstraint/Point2PointConstraint.

I was wondering if there's a specific reason for this, or should I be using the TranslationalLimitMotor/RotationalLimitMotor's m_accumulatedImpulse instead for 6DofConstraint ? In that case there doesn't seem to be any methods to get these from the constraint class itself.

If anyone could shed some light on this, I would really appreciate it.

Purpose:
I need this to be accessed in a tickcallback to check to see how much impulse was applied to a constraint to keep it in place...and then make some choices whether to break the constraint or not.

Thanks,
Ram.

[Mathieu Colleaux]

Hi,

I am experiencing the same issue. I'am trying to get the applied impulse from a btTypedConstraint to decide if I need to break the constraint or not, but the value returned by getAppliedImpulse() is always zero. I had a look at the solver and I didn't saw any update of the m_appliedImpulse of the constraint. Maybe there is an issue here.

Anyway, if you have any information that could help I would appreciate this.

Cheers,

M. Colleaux

[ramsampath]

Yes, I was able to get it to work with modifications to bullet.

In bullet 2.70 source code

under the directory
/src/BulletDynamics/ConstraintSolver

and filename

btGeneric6DofConstraint.cpp

I changed the following line

Code: Select all

    m_linearLimits.solveLinearAxis(
                m_timeStep,
                jacDiagABInv,
                m_rbA,pointInA,
                m_rbB,pointInB,
                i,linear_axis, m_AnchorPos);

to

Code: Select all

    m_appliedImpulse += m_linearLimits.solveLinearAxis(
                m_timeStep,
                jacDiagABInv,
                m_rbA,pointInA,
                m_rbB,pointInB,
                i,linear_axis, m_AnchorPos);

        }
    }

and the following line

Code: Select all

  m_angularLimits[i].solveAngularLimits(m_timeStep,angular_axis,angularJacDiagABInv, &m_rbA,&m_rbB);

to

Code: Select all

   m_appliedImpulse +=  m_angularLimits[i].solveAngularLimits(m_timeStep,angular_axis,angularJacDiagABInv, &m_rbA,&m_rbB);

This is probably fixed in later bullet versions...

so maybe you want to do something similar in btTypedConstraint.cpp

Hope this helps you.

Ram.

[kate]

Hi Mathieu,

Did you find a way to get this to work? (I'm trying to do the same thing).

Thanks,

Kate

[Mathieu Colleaux]

Hi Kate,

I decided to add an impulse accumulator in btTypedConstraint :

Code: Select all

struct btImpulseAccumulator {
		btVector3 m_linImpulseA;
		btVector3 m_angImpulseA;

		btVector3 m_linImpulseB;
		btVector3 m_angImpulseB;
	};

It stores the impulses applied by the constraint solver and it is reseted every simulation step (at the begining of the buildJacobian method).
You will need to modify the constraint solver to store the impulses correctly.

Once the constraint is solved I compare the square length of the impulses with a threshold to determine if the constraint is broken or not.

Code: Select all

void updateImpulseBreakable (btScalar timestep)
	{
		// A broken constraint should have been already removed from the simulation
		btAssert(!m_broken);

		if (isBreakable())
		{
			float length1, length2;
			// Compute impulses from velocities
			m_accumulator.m_linImpulseA *= timestep;
			m_accumulator.m_angImpulseA *= timestep;
			m_accumulator.m_linImpulseB *= timestep;
			m_accumulator.m_angImpulseB *= timestep;

			length1 = m_accumulator.m_linImpulseA.length2() + m_accumulator.m_angImpulseA.length2();
			m_broken |= length1 > (m_BreakThreshold*m_BreakThreshold);

			length2 = m_accumulator.m_linImpulseB.length2() + m_accumulator.m_angImpulseB.length2();
			m_broken |= length2 > (m_BreakThreshold*m_BreakThreshold);

			if (m_broken)
			{
				// Fire a callback to inform the user that the constraint have been broken during the current frame
				if (gConstraintBrokenCallback)
				{
					float maxImpulse = (length1 > length2) ? length1 : length2;
					gConstraintBrokenCallback(*this, btSqrt(maxImpulse));
				}
				m_broken=true;
			}
		}
	}

If the constraint is broken it will be removed at the begining of the next simulation step.

I hope this helps,

Mathieu

[Erwin Coumans]

Hi Mathieu,

It would be great if you could provide a patch for this, and the breakable constraint callback, against latest trunk. Would that be possible?

UPDATE: We'll try to fix appliedImpulse for joints before final 2.75 release: http://code.google.com/p/bullet/issues/detail?id=227

Thanks,
Erwin

[kate]

Cool, thanks Mathieu - I'll give that a go.

K.

[kate]

Hi Erwin,

UPDATE: We'll try to fix appliedImpulse for joints before final 2.75 release: http://code.google.com/p/bullet/issues/detail?id=227

That would be really cool - thanks!

Kate