From: owner-3dui@hitl.washington.edu on behalf of Chad Wingrave
[cwingrav@vt.edu]
Sent: Saturday, December 16, 2000 12:49 PM
To: 3dui List
Subject: Describing Interfaces ... ex. Ray Casting


Something I have been thinking about:  Since interaction is generally a
cause and effect, if we explain things in the sense of triggers and
actions, we might be able to explain interaction simply and hopefully with
similar expressions.

The environment might have a set of selectable objects and a set of those
that are selected.  In this simple case to follow, I only allow one item
to be selected at a time but that could easily be changed.

Additionally, there are very few structures that need to be described for
interactions in VEs.  Something to specify location and orientation
(position) and an object (contains a position and probably other things
which interaction may not need to know about) are about it.  For most
interaction techniques, the functions that operate on data are common and
after describing one interaction, another could be described hopefully
with many of the same parts (what's this?  a vocabulary?).  For example,
Occlusion Selection and Ray Casting have many of the same qualities.  GoGo
too compares nicely and in my Ray Casting example it is easy to see a case
where the function PointingError could be used for Occlusion and GoGo.

I do agree that written descriptions are nice but a more formal
description would be additionally needed as it allows researchers
to converse using the same vocabulary and hopefully describe new
interactions in terms of others.  Hopefully VEs could even be specified in
this language.   Also, since interaction is generally an event based
system, it should be defined as one with initialization stages, triggers
and actions.


The following is a example of RayCasting.  Comments, criticisms and
corrections are welcome.

	-Chad Wingrave
	 Virginia Tech CS

Init: 
// ----------------------------------------------------------

set of objects highlighted = empty
locatordevice hand = tracked
int minerror       = .02  (or whatever is appropriate)

Events:
// ----------------------------------------------------------

trigger: button, pinch, etc.
action : Select(highlighted)


trigger: hand changed location
action : Set(highlighted,
          RaycastSelect(Position(hand),SelectableObjects(),minerror))


Implementation:
// ----------------------------------------------------------

// finds error between all selectable objects and if the most
//   likely selected is above a threshold, select it
RaycastSelect(_pointer: position, _objects : set of objects,
              _raycastminerror: real )
{
  min = 1.0, retval = NULL
  foreach obj in _objects
    temp = RaycastError(Pointer,Obj)
    if ( temp < min ) min = temp, retval = obj
  if ( min > _raycastminerror ) retval = NULL;
  return retval
}

// Gives an error value for the ray casting.  Normalizes between 1 and 0
RaycastError(_pointer: position, _object : object) : real
{  return PointingError(_pointer,Position(_object)) / 180  }


// Should give the amount that the ray is off of the orientation.
//   NOTE: My math most likely is incorrect but you get the point
PointingError(_ray : position, _point : position) : angle
{
  i = Invert(_ray)*_point
  return  2*Sqr(X(i)) / (2*X(i)*Sqrt(Sqr(X(i))+Sqr(Z(i))))
}