[ros-users] PropagationDistanceField usage?

[Matthew Jacob Klingensmith]

Hello ROS users list,

I am currently trying to develop an application using the
PropagationDistanceField class found in the distance_field package.
However, I am getting some unexpected results when I do simple tests with
points.

Constructing the distance field with parameters:

x_size: 5.0 meters
y_size: 5.0 meters
z_size: 5.0 meters
origin_x : -2.5 meters
origin_y : -2.5 meters
origin_z : -2.5 meters
resolution: 0.05 meters
max_distance: 10 meters

And doing the following:
////

distanceField->reset();
std::vector<btVector3> points;
for(double y = -2.5; y < 2.5; y+=0.05)
{
    points.add(btVector3(0,y,0));
}

distanceField->addPointsToField(points);

////

I should expect to get a distance field which has a line of zeroes along
the y axis, with increasing distances from there. However, the distance
field I actually get has a line of zeroes along a line roughly where y=0,
and both x and z are set to what should be y.

Here is a screenshot of the vizualization I get out of rviz when I try
adding in a line of points along the y axis (I added the colored markers
along the XY plane myself, the red markers come out of the vizualization
method):

http://img155.imageshack.us/img155/2771/yaxis.png

When I instead try to add a line along the x or z axes, I get an even
stranger result, where there is simply a point at the origin and nothing
else:

http://img545.imageshack.us/img545/6683/xaxis.png

Seeing these results, I am wondering if I am using the distance field
class properly. I was under the impression that "addPointsToField" was
supposed to take in a vector of points in world coordinates, and then
compute the distance transformation of the voxel field given those points.
Is this correct?

Thank you

Matthew Klingensmith
mklingen at andrew.cmu.edu