Possible flying vehicle improvement?

I have been digging around in the flying vehicle code to see if I could resolve the annoying tendancy of certain flying vehicles to either not hover in one place when the are stationary at hover height, drift off erratically on their own or sink into terrain/interiors when left to their own devices.

I think I have found the problem in flyingVehicle.cc but thought i would post here first instead of bugs to see if my fix causes any other issues. The code referred to below is still the same in the 1.4 CVS code.


The problem is inside FlyingVehicle::updateForces. In the current code, force is first initialised with gravity acting on the vehicle in the z direction:-

...
Point3F force  = Point3F(0, 0, sFlyingVehicleGravity * mRigid.mass * mGravityMod);
...

This seems a very reasonable starting point since we want our flying vehicle to fall naturally under the influence of gravity BUT, later on in the function, we take the vehicles orientation into account (using zv from getTransform) when applying the hover forces :-

"...
   // Hovering Jet
   F32 vf = -sFlyingVehicleGravity * mRigid.mass * mGravityMod;
   F32 h  = getHeight();
   if (h <= 1) {
      if (h > 0) {
         vf -= vf * h * 0.1;
      } else {
         vf += mDataBlock->jetForce * -h;
      }
   }
   force += zv * vf;

// Damping "surfaces"
..."

Since the vehicle is always at some orientation other than perfectly aligned up/down, this applied hover force includes x and y components and does not therefore cancel out the vertical only gravity force we started out with. The result is that it nevers hovers in one place as I think was the intention of the hovering jet code but moves around while it tries in vain to cancel out gravity (and generally does all sorts of strange things).

To fix this, first you need to set up the initial force variable to include gravity in the same manner that the 'anti gravity' force for hover is calculated later in the code.


At the beginning of FlyingVehicle::updateForces where force is initialised :-

"...
//Point3F force  = Point3F(0, 0, sFlyingVehicleGravity * mRigid.mass * mGravityMod); //Hover fix
Point3F force  = Point3F(0, 0, 0);
F32 vf = -sFlyingVehicleGravity * mRigid.mass * mGravityMod; //vf def moved to here
force += zv * -vf;  //gravity force now in relation to orientation
..."

This change sets up the initial gravity force to take into account orientation. I have moved the definition of vf to this point in the code so I can use it here as well as later in the function. The original definition of force is commented out in the snippet above and the new lines added immediately after.


Further down FlyingVehicle::updateForces, we change the Hover Jet part of the function.

The new code looks like this :-

"...
   // Hovering Jet
   F32 h  = getHeight();
   if (h <= 1) { //true if we are hovering
      if (h > 0) {
         vf -= vf * h * 0.1;
      } else {
         vf += mDataBlock->jetForce * -h;
      }
   force += zv * vf; //cancel oriented gravity, scaled to hover
   }
   else //we are not hovering so change gravity back to vertical
   {
   force += zv * vf; //cancel oriented gravity
   force +=Point3F(0,0,-vf); //add z component of gravity only
   }

   // Damping "surfaces"
..."

In the snippet above if we are hovering we apply an oriented hover force to cancel out our initial
oriented gravity. If we are not hovering, we remove our oriented gravity and apply vertical gravity instead. Notice that the braces are now slightly different from the original code.

With this change, flying vehicles hover in one place when at their hover height and fall under gravity when above hover height. I have noticed less of a tendancy for the vehicle to disappear into terrain and/or surfaces after this fix but it will still do it if you try hard enough.

Now, if only someone could fix the collision freezing as well :)