A Cliff Detection - JASS

08-26-2007, 12:57 PM	#1
Silvenon	Hello, I'm trying to detect if a collision missile hit the cliff, because I want it to bounce from it. This is very complicated, I know. The bounce angle, cliff detection and everything. I'm open to complicated solutions because I don't think a simple one exists. Thanks in advance.

08-26-2007, 02:04 PM	#2
DioD	terrain normal function from any vector system

08-26-2007, 02:36 PM

Alexander244

Terrain normal vector can be obtained from the cross product of two vectors placed on the terrain at the point of interest.

Something like this:

Cross Product:

static method CrossProduct takes vector v1, vector v2 returns vector
  local real x = (v1.y*v2.z-v1.z*v2.y)
  local real y = (v1.z*v2.x-v1.x*v2.z)
  local real z = (v1.x*v2.y-v1.y*v2.x)
  return vector.create(x, y, z)
endmethod

Terrain Normal:

static method GetTerrainNormal takes real x, real y, real radius returns vector
  local vector x_vec
  local vector y_vec
  local vector normal
  local real z

  call MoveLocation(TempLoc, x - radius, y)
  set z = GetLocationZ(TempLoc)
  call MoveLocation(TempLoc, x + radius, y)
  set x_vec = vector.create(2 * radius, 0.0, GetLocationZ(TempLoc) - z)

  call MoveLocation(TempLoc, x, y - radius)
  set z = GetLocationZ(TempLoc)
  call MoveLocation(TempLoc, x, y + radius)
  set y_vec = vector.create(0.0, 2 * radius, GetLocationZ(TempLoc) - z)

  set result = vector.CrossProduct(x_vec, y_vec)

  call x_vec.destroy()
  call y_vec.destroy()

  return normal
endmethod

(I am assuming you know vJass well enough to be able to fill in the gaps - like making a vector struct etc.)

Cliffs have a 45 degree angle, but visually appear to have a 90 degree angle. You will need to find some way of detecting if the angle is 45 degrees and the terrain is non-pathable etc. then remove the z-component from the terrain normal vector.

To bounce you could do something like rotating the vector of the missile around the terrain normal, and then amplifying it by -1.

I am not sure if the above are the best methods/work... the best way to go is probably to look through some vector systems, to see how these problems were solved.