[ODE] Laser range finder

[Shamyl Zakariya]

Also, depending on how "real world" you want your laser range finder to 
behave you might want to take into account the angle which the ray hits 
whatever it hits.

What I mean is, real range finders tend to get less accurate as the 
intersection angle goes away from perpendicular.

In my simulator, I use this angle to add a random error quotient, so 
the reported distance gets noisier the farther from perpendicular the 
intersection becomes.

	dContactGeom contact;	
	if ( dCollide( _rayGeom, geom, 1, &contact, sizeof(dContactGeom)) > 0 )
	{
		if ( contact.depth < _distance )
		{
			_contactPos[0] = contact.pos[0];
			_contactPos[1] = contact.pos[1];
			_contactPos[2] = contact.pos[2];
			_contactPos[3] = contact.pos[3];

			_distance = contact.depth;
			
			/*
				now calculate dot product of ray to intersection against
				the normal of the surface ray intersects. This will be
				used to influence the error calculations.				
			*/
			
			dVector3 start, dir;
			dGeomRayGet( _rayGeom, start, dir );

			dVector3 toStart = { start[0] - _contactPos[0],
			                     start[1] - _contactPos[1],
								 start[2] - _contactPos[2] };

			normalize( toStart );
			_contactDotProduct = dDOT( toStart, contact.normal );
		}
	}


Then, later on when I query the distance,


	/*
		otherwise, we need to calculate a psuedorandom error
		First, generate an error quotient from -1 to 1, influenced
		by incidence of ray intersection to target. Where the error
		increases as the ray intersection deviates from perpendicular.
	*/
	dReal error = (( 1.0 - _contactDotProduct ) * 2 *
	              ( ((dReal) (random() % 1000)) / 1000.0 )) - 0.5;

	/*
		Now scale the error up based on how far away the intersection is.
	*/
	
	dist += error * ( dist / max ) * _error;
	if ( dist >= max ) dist = max;

Also, since the range finders I'm emulating have a max range, I cap it 
if it's too far.


Anyhow, I hope this helps...


On Sep 19, 2004, at 6:34 PM, John Miles wrote:

>
>> To paraphrase the documentation (did you search for "ray" ?)
>>
>> 1) create a dRay geom
>> 2) set the position and direction appropriately using the dRay 
>> functions
>> 3) set the length to be the total size of your world
>> 4) run dSpaceCollide2() to collide your ray against the world
>> 5) the contact you get back from within dCollide has a "depth"
>> value which is the length of the ray to the contact -- remember,
>> and use, the smallest depth returned
>
> FYI, be careful using really-long rays with the dTerrainY/Z 
> contributions.
> If you fire a ray at an angle that the terrain collider doesn't 
> trivially
> reject (e.g., 45 degrees from just above the center of a world to a 
> point
> beneath one of its corners) , performance suffers greatly compared to 
> an
> orthogonal ray, or a shorter one.
>
> Or, that may be a general property of the quadtree space, for all I 
> know.  I
> didn't investigate.
>
> The other thing about rays is that they "miss" trimeshes if they 
> happen to
> strike an edge.
>
> -- jm
>
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>