btHeightfieldTerrainShape giving a flat terrain

[dumbo2007]

Hi,


I modified the TerrainDemo example to load my own raw values. I am trying to generate raw sine wave values and generate the sine wavy terrain from these values. Here are the modified sections :

An extra case called case eFile: was added to getRawHeightfieldData

Code: Select all

static byte_t *
getRawHeightfieldData
(
eTerrainModel model,
PHY_ScalarType type,
btScalar& minHeight,
btScalar& maxHeight
)
{
//	std::cerr << "\nRegenerating terrain\n";
//	std::cerr << "  model = " << model << "\n";
//	std::cerr << "  type = " << type << "\n";

	long nElements = ((long) s_gridSize) * s_gridSize;
//	std::cerr << "  nElements = " << nElements << "\n";

	int bytesPerElement = getByteSize(type);
//	std::cerr << "  bytesPerElement = " << bytesPerElement << "\n";
	btAssert(bytesPerElement > 0 && "bad bytes per element");

	long nBytes = nElements * bytesPerElement;
//	std::cerr << "  nBytes = " << nBytes << "\n";
	byte_t * raw = new byte_t[nBytes];
	btAssert(raw && "out of memory");

	// reseed randomization every 30 seconds
//	srand(time(NULL) / 30);

	// populate based on model
	switch (model) {
	case eRadial:
		setRadial(raw, bytesPerElement, type);
		break;

	case eFractal:
		for (int i = 0; i < nBytes; i++)
		{
			raw[i] = 0;
		}
		setFractal(raw, bytesPerElement, type, s_gridSize - 1);
		break;

	case eFile:
		for (int i = 0; i < nBytes; i++)
		{
			raw[i] = 20.f + 20.f*sin(float(i));
		}		
		break;

	default:
		btAssert(!"bad model type");
	}

	if (0) {
		// inside if(0) so it keeps compiling but isn't
		// 	exercised and doesn't cause warnings
//		std::cerr << "final grid:\n";
		dumpGrid(raw, bytesPerElement, type, s_gridSize - 1);
	}

	// find min/max
	for (int i = 0; i < s_gridSize; ++i) {
		for (int j = 0; j < s_gridSize; ++j) {
			float z = getGridHeight(raw, i, j, type);
			printf("i=%d, j=%d,  z=%f\n" , i, j, z);

			// update min/max
			if (!i && !j) {
				minHeight = z;
				maxHeight = z;
			} else {
				if (z < minHeight) {
					minHeight = z;
				}
				if (z > maxHeight) {
					maxHeight = z;
				}
			}
		}
	}

	if (maxHeight < -minHeight) {
		maxHeight = -minHeight;
	}
	if (minHeight > -maxHeight) {
		minHeight = -maxHeight;
	}

//	std::cerr << "  minHeight = " << minHeight << "\n";
//	std::cerr << "  maxHeight = " << maxHeight << "\n";

	return raw;
}

I also modified the initialization routine :

Code: Select all

void TerrainDemo::initialize(void)
{
//	std::cerr << "initializing...\n";

	// set up basic state
	m_upAxis = 1;		// start with Y-axis as "up"
	m_type = PHY_FLOAT;
	m_model = eFile;//eFractal;
	m_isDynamic = true;

	// set up the physics world
	m_collisionConfiguration = new btDefaultCollisionConfiguration();
	m_dispatcher = new btCollisionDispatcher(m_collisionConfiguration);
	btVector3 worldMin(-1000,-1000,-1000);
	btVector3 worldMax(1000,1000,1000);
	m_overlappingPairCache = new btAxisSweep3(worldMin,worldMax);
	m_constraintSolver = new btSequentialImpulseConstraintSolver();
	m_dynamicsWorld = new btDiscreteDynamicsWorld(m_dispatcher,m_overlappingPairCache,m_constraintSolver,m_collisionConfiguration);

	// initialize axis- or type-dependent physics from here
	this->resetPhysics();
}

and the grid size is set to a lower value :

Code: Select all

static const int s_gridSize			= 8 + 1;  // must be (2^N) + 1

Resetting of physics is the same, for reference here it is :

Code: Select all

/// called whenever key terrain attribute is changed
void TerrainDemo::resetPhysics(void)
{
	// remove old heightfield
	clearWorld();

	// reset gravity to point in appropriate direction
	m_dynamicsWorld->setGravity(getUpVector(m_upAxis, 0.0, -s_gravity));

	// get new heightfield of appropriate type
	m_rawHeightfieldData =
	    getRawHeightfieldData(m_model, m_type, m_minHeight, m_maxHeight);
	btAssert(m_rawHeightfieldData && "failed to create raw heightfield");

	bool flipQuadEdges = false;
	btHeightfieldTerrainShape * heightfieldShape =
	    new btHeightfieldTerrainShape(s_gridSize, s_gridSize,
					  m_rawHeightfieldData,
					  s_gridHeightScale,
					  m_minHeight, m_maxHeight,
					  m_upAxis, m_type, flipQuadEdges);
	btAssert(heightfieldShape && "null heightfield");

	// scale the shape
	btVector3 localScaling = getUpVector(m_upAxis, s_gridSpacing, 1.0);
	heightfieldShape->setLocalScaling(localScaling);

	// stash this shape away
	m_collisionShapes.push_back(heightfieldShape);

	// set origin to middle of heightfield
	btTransform tr;
	tr.setIdentity();
	tr.setOrigin(btVector3(0,-20,0));

	// create ground object
	float mass = 0.0;
	localCreateRigidBody(mass, tr, heightfieldShape);
}

I am directly setting the raw values using raw = 20.f + 20.f*sin(float(i));
But this gives me only a flat terrain. It seems that something is wrong in the values I set in raw[] as this section in getRawHeightfieldData() always gives z=0 and consequently m_minHeight and m_maxHeight are both 0 in resetPhysics()

Code: Select all

// find min/max
	for (int i = 0; i < s_gridSize; ++i) {
		for (int j = 0; j < s_gridSize; ++j) {
			float z = getGridHeight(raw, i, j, type);
			printf("i=%d, j=%d,  z=%f\n" , i, j, z);

			// update min/max
			if (!i && !j) {
				minHeight = z;
				maxHeight = z;
			} else {
				if (z < minHeight) {
					minHeight = z;
				}
				if (z > maxHeight) {
					maxHeight = z;
				}
			}
		}
	}

There are functions in the demo which generate fractal and radial heightfields. I cannot understand how differently they are setting the raw values so that the m_minHeight and m_maxHeight are reported correctly. Do I need to align the values I set in raw[] in someway ?

Thanks

[dumbo2007]

Any idea about this anyone....has anyone tried reading heightmap values or modifying the TerrainDemo to read in heightmap values ?

[Dave Driesen]

Hi,

In your getRawHeightfieldData function, your data has an incorrect alignment:

Code: Select all

case eFile:
      for (int i = 0; i < nBytes; i++)
      {
         raw[i] = 20.f + 20.f*sin(float(i));
      }      
      break;

You are writing to raw which will be byte-aligned; raw is declared as byte_t*.
Use the convertFromFloat function, and correct your alignment (multiplying i by bytesPerElement, or alternatively changing your loop-expression).

Try this:

Code: Select all

case eFile:
      for (int i = 0; i < nElements; i++) {
		// simple triangle waveform
		float z=(i % 2) * 3.0f;
		convertFromFloat(&raw[i*bytesPerElement], z, type);
	}

Doesn't use your sine wave as I'm not sure of the actual effect you desire, but it does show you how to fill up the heightfield data.

Happy coding,
Dave

[dumbo2007]

Thanks Dave. Yes it work now