The C-Wing of LSRC is an extremely challenging environment with many thin vertical objects and many partially reflective/transmissive objects.  We have included an annotated map below.  Please scroll down for more maps and experimental details. Click to enlarge the map below.

We have taken pictures of the areas indicated with red numbers above:

  1. Starting position (human perspective)
  2. Catwalk (robot perspective)
  3. Hallway with cart (robot perspective)
  4. Turn with glass wall (robot perspective)
  5. View of lab space (human perspective) (90 degrees left from robot orientation)
  6. Loop closing (human perspective)
  7. Loop closing (robot perspective)

Understanding our Maps

The maps we have shown here are the maps associated with the highest probability particle.  White areas have never been scanned by the laser (unknown).  Darker shades of gray indicate higher probability of stopping the laser (greater opacity).  DP-SLAM 1.0 maps are an exception.  DP-SLAM 1.0 did not distinguish between empty space unknown space and had a deterministic model of laser penetration, so 1.0 maps are black and white.

Some maps included below, typically the bad ones, have color. This is diagnostic information printed by DP-SLAM 2.0 during a run.  The red pixels in the image indicate the endpoints of laser scans which did not match well with the map. These points tend to correspond either to new observations which are about to be entered into the map, or else discrepencies between the map and the observations. The blue points correspond to laser endpoints which match well with the map; i.e., the robot already believes that there is an object in the square where this observation terminates. The cyan pixels indicate observations where the sensor was unable to see an object before reaching the maximum effective range of 8m.

Log File
Grid Resolution
Maps (Click to enlarge) Comments
Loop 25

Loop 25 DP-SLAM 2.0


Lower quality map.
Loop 25 DP-SLAM 2.0
(simple laser model)

Loop does not close
properly with simple
laser model.
Loop 25

The deterministic model
cannot handle the balusters
of the catwalk railing.