Weak hinge motor

[dmp32]

Hello,
I'm building a simulator for a 4-legged robot using Bullet, every leg has 2 hinge constraints with motors (simulating servos).

Called at each step :

Code: Select all

        btScalar velocity(1.0), maximpulse(10.0);
        btScalar diff= phinge->getHingeAngle() - target_angle;

        if(diff < -epsilon)
                phinge->enableAngularMotor(true, velocity, maximpulse);
        else if(diff > epsilon)
                phinge->enableAngularMotor(true, -velocity, maximpulse);
        else
                phinge->enableAngularMotor(false, 0.0, 0.0);

However, the motor is unable to lift the 0.340-kilogram, 20cm-wide robot until I increase the velocity to say 10.0, but then it's moving too fast.
I simply can't simulate a slow but high-torque motor.

Please help me, I don't understand what's the problem here...
Thanks.

[BinhNguyen]

I was just working on a similar problem today. I couldn't get a joint to bend slowly by adjusting the target velocity and the maximum impulse. When the target velocity is a small number, the joint would not bend at all no matter how big the maximum impulse was.

Instead I made the target angle vary slowly over time. I also had to constantly activate the rigid bodies or else they went to sleep.

That reminds me, perhaps all I have to do is constantly activate the rigid bodies when the target velocity is small. I'll have to try that tomorrow.

[dmp32]

This looks like a bug to me. Isn't it ?
However, in the ForkLift demo, there is a hinge with a very small velocity (0.1), and it works fine with lifting a 4-kilogram load.
But all my attempts to reproduce it have failed so far...
Disabling deactivation doesn't help.

Edit :
I just tried the following code (which updates the target angle dynamically) :

Code: Select all

        diff= h->getHingeAngle() - targetangle;
        h->enableAngularMotor(true, 0.0, motortorque);
        if(diff < -epsilon)
            h->setMotorTarget(h->getHingeAngle() + motorspeed * timeStep, timeStep);
        else if(diff > epsilon)
            h->setMotorTarget(h->getHingeAngle() - motorspeed * timeStep, timeStep);
        else
            h->enableAngularMotor(false, 0.0, 0.0);

But the result is the same : the motor is still too weak when its speed is low, even with high impulse.
I can't find the origin of the problem.

Please someone, save my neurons xD !

[dmp32]

Here is another test :

[BinhNguyen]

I tried updating the target angle dynamically like you did and it did not work for me either. I suspect that the values are too small and are lost in the calculations.

Instead, I kept my own target angle variable and increased it slowly overtime. This approach worked.

Code: Select all

void Update( float deltaTime )
{
	alphaRigidBody->activate();
	bravoRigidBody->activate();

	bool isEnableMotor = true;
	btScalar maxMotorImpulse = 1.0f; // 1.0f / 8.0f is about the minimum

	hingeConstraint->enableMotor( isEnableMotor );
	hingeConstraint->setMaxMotorImpulse( maxMotorImpulse );
	
	targetAngle += 0.1f * deltaTime;
	hingeConstraint->setMotorTarget( targetAngle, deltaTime );	
}

[dmp32]

Yep, this approach actually works ! I figured it out just before reading your response.

Code: Select all

        //setMaxMotorImpulse and enableMotor are called at the creation of the hinge

        h= hinges[i];
        diff= mcurangles[i] - mtargetangles[i];
        if(diff < -motorspeed * timeStep)
            mcurangles[i] += motorspeed * timeStep;
        else if(diff > motorspeed * timeStep)
            mcurangles[i] -= motorspeed * timeStep;
        else
            mcurangles[i]= mtargetangles[i];

        h->setMotorTarget(mcurangles[i], timeStep);

Unfortunately, I don't have enough knowledge to write a patch for enableAngularMotor...

[BinhNguyen]

I'm glad you got your program to work!

Just in case you might find it useful, I was experimenting with btGeneric6DofConstraint and I found that it worked with small velocities without having to use a workaround.