How to do water?

[pHySiQuE]

Now that the buoyancy force function is gone, how do we make a water plane? I have this function but it seems to add random force to the object and doesn't look anything like buoyancy forces:

Code: Select all

   void NewtonDynamicsBody::AddBuoyancyForce(const float nx, const float ny, const float nz, const float d, const float fluidDensity, const float fluidViscosity)
   {
      if (collision)
      {
         dFloat shapeVolume = NewtonConvexCollisionCalculateVolume(collision);
         if (shapeVolume > 0.0f)
         {
            dFloat shapeOrigin[3] = { 0.0, 0.0, 0.0 };
            dFloat matrix[16];
            dFloat gravityVector[3] = { world->gravity.x, world->gravity.y, world->gravity.z };
            dFloat accel[3];
            dFloat alpha[3];
            dFloat fluidPlane[4] = { nx, ny, nz, d };
            
            NewtonBodyGetMatrix(body,matrix);
            NewtonBodyGetCentreOfMass(body,shapeOrigin);

            float waterToSolidVolumeRatio=fluidDensity;
            
            dFloat dens = 1.0f / (waterToSolidVolumeRatio * shapeVolume);

            NewtonConvexCollisionCalculateBuoyancyAcceleration(collision, matrix, shapeOrigin, gravityVector, fluidPlane, fluidDensity, fluidViscosity, accel, alpha);
            if (accel[0] != 0.0 || accel[1] != 0.0 || accel[2] != 0.0 || alpha[0] != 0.0 || alpha[1] != 0.0 || alpha[2] != 0.0)
            {
               NewtonBodyAddForce(body, accel);
               NewtonBodyAddTorque(body, alpha);
            }
         }
      }
   }

[Julio Jerez]

the function is this

Code: Select all

void NewtonConvexCollisionCalculateBuoyancyAcceleration (const NewtonCollision* const convexCollision, const dFloat* const matrix, const dFloat* const shapeOrigin, const dFloat* const gravityVector, const dFloat* const fluidPlane, dFloat fluidDensity, dFloat fluidViscosity, dFloat* const accel, dFloat* const alpha)

the demo is ArchimedesBuoyancy.cpp

[pHySiQuE]

This is what my behavior looks like (video freezes until about halfway through, just watch the whole thing):


Is this right?

[Julio Jerez]

I first try making a box with a more regular aspect ratio.
see if you can get I stable first and them start change the dimensions gradually.

essentially the alchemies buoyancy force is a function of the volume under the plane
a shape with a very skew volume lead to a very stiff volume rate of change.

you can adjust it with damping, to a certain limit, beyond that the only thong that can make stable is a higher frequency update so the volume become a more smooth function.

If you can run the SDK demo, try to place one of the boxes with the same shape size, make I can see how to adjust it to be more stable.

[pHySiQuE]

I didn't really notice any effect with fluid viscosity. Shouldn't the viscosity of the water cause a huge amount of damping? Right now it just looks like it's been dropped a vertical wind tunnel:

[Julio Jerez]

after I changed the function to operate on shapes rather than on a rigid body, the viscosity is not longer used. That was delegate to the end application

For realistic calculation the effect for the fluid viscosity on the body, the velocity must be known, and realistic viscose effects is a trivial calculation because viscosity act like a non linear thing on objects, It is like fly lift forces calculation, you can do a trivial calculation ot you can do a surface integration. You can however fake it very easy, in you function you can do something like this.

Code: Select all

            dMatrix matrix;
            dVector cog;
            dVector accelPerUnitMass;
            dVector torquePerUnitMass;
            const dVector gravity (0.0f, DEMO_GRAVITY, 0.0f, 0.0f);

            NewtonBodyGetMatrix (visitor, &matrix[0][0]);
            NewtonBodyGetCentreOfMass(visitor, &cog[0]);
            cog = matrix.TransformVector (cog);
            NewtonCollision* const collision = NewtonBodyGetCollision(visitor);

            
            dFloat shapeVolume = NewtonConvexCollisionCalculateVolume (collision);
            dFloat fluidDentity = 1.0f / (m_waterToSolidVolumeRatio * shapeVolume);
            dFloat viscosity = 0.995f;

            NewtonConvexCollisionCalculateBuoyancyAcceleration (collision, &matrix[0][0], &cog[0], &gravity[0], &m_plane[0], fluidDentity, viscosity, &accelPerUnitMass[0], &torquePerUnitMass[0]);

            dVector force (accelPerUnitMass.Scale (mass));
            dVector torque (torquePerUnitMass.Scale (mass));

// here this is show it was done before in 1.5 and 2.xx
            dVector omega;
            NewtonBodyGetOmega(visitor, &omega[0]);
            omega = omega.Scale (viscosity);
            NewtonBodySetOmega(visitor, &omega[0]);

            NewtonBodyAddForce (visitor, &force[0]);
            NewtonBodyAddTorque (visitor, &torque[0]);


I added that to the demo now.

[pHySiQuE]

The damping made the behavior much better.