likelihood field of map

asked 2012-12-05 12:13:17 -0500

721 ●54 ●64 ●71

updated 2012-12-05 12:38:18 -0500

Hi, I'd like to ask about a convert of a map data.

Amcl package uses map.pgm with map.yaml.
However, map.png which is gained by slam_gmapping
has solid edges to represent obstacles(walls, tables and so on).

When and where a map data is converted to a likelihood field ?
(If possible, could you show me the correspond part of source codes?)
And which parameter is used to determine the likelihood field's standard deviation?

Thanks in advance.

answered 2012-12-06 04:00:45 -0500

Ivan Dryanovski

4954 ●52 ●70 ●86 http://robotics.ccny.cuny.edu/

updated 2012-12-07 12:35:12 -0500

As far as I know, gmapping maintains occupancy values internally. However, the ROS wrapper thresholds them into binary free/occupied values.

Take a look at this file:

slam_gmapping.cpp

Starting at line 625:

for(int x=0; x < smap.getMapSizeX(); x++)
{
  for(int y=0; y < smap.getMapSizeY(); y++)
  {
    GMapping::IntPoint p(x, y);
    double occ=smap.cell(p);
    assert(occ <= 1.0);
    if(occ < 0)
      map_.map.data[MAP_IDX(map_.map.info.width, x, y)] = -1;
    else if(occ > occ_thresh_)
    {
      //map_.map.data[MAP_IDX(map_.map.info.width, x, y)] = (int)round(occ*100.0);
      map_.map.data[MAP_IDX(map_.map.info.width, x, y)] = 100;
    }
    else
      map_.map.data[MAP_IDX(map_.map.info.width, x, y)] = 0;
  }
}

The default value for occ_thresh_ is 0.25. It appears that if you want the occupancy values, you can simply revert the line which was commented out, and change the code to:

map_.map.data[MAP_IDX(map_.map.info.width, x, y)] = (int)round(occ*100.0);
//map_.map.data[MAP_IDX(map_.map.info.width, x, y)] = 100;

I have created a ticket here.

EDIT: response to comment

I'm not sure whether amcl currently assumes binary values or not.

With laser sensors, you typically reach saturation probabilities (0 or 1) pretty fast, since lasers have a small uncertainty spread. My intuition is that since the nav stack is mainly developed for laser-equipped ground robots, the binary value is a good approximation.

edit flag offensive delete link

Comments

Thank you for your kind answer and creating the ticket !! After I change the line, I need to modify amcl package to use likelihood field which doesn't have binary values but 0～255 values don't I ? Why do you think nav stack use binary values even though it's less robust ?

moyashi ( 2012-12-06 11:59:49 -0500 )edit

Thank you for your response.I'd like to be sure whether amcl assumes binary or not. Would you mind if I ask the question in an additional question form which includes url info of this form?

moyashi ( 2012-12-09 13:23:33 -0500 )edit

sure, sounds like a good idea