Final project information

• About the data file format
• Conversion script and standard data files
• Pictures

• Units: centimeters versus meters, degrees versus radians

• Different number of laser readings (181 for "stanford", 180 for the others)

• Different representation of laser readings (lines versus ranges)

• The "aces" file (which uses lines) indicates the offset from the center of the laser range scanner from the robot center.

• While the "aces" file nicely alternates the position and laser readings, the "intel" file generally has two position lines in a row (which are in general different), and the "stanford" file generally has a number of laser scans in a row (evidently without any position information associated with the scans).

I've settled on a data format which will:

• Alternate position and laser scans, starting with a position datum. I am doing this by:
  • discarding any laser scan before the first position datum
  • only using the last position datum before a laser scan
  • only using the first laser scan after a position datum

  I think these are reasonable choices, but you can let me know if you think otherwise.

• Distances will be in meters (to 3 decimal places), angles will be in degrees (to one decimal place). The choice of units for angles follows my general convention of making external representations of angles (than people may read or look at) in degrees, while all internal representations (i.e. only seen by the program) are in radians.

• Laser scans will be represented by ranges (not lines)

• There will be a short header containing the following:

  # laser-offset 0.06 0.0 0.0
  # laser-angles 180 -89.5 89.5 1.0
  

  This will tell you the offset (which I have set to 0.0 0.0 0.0 for the "intel" and "stanford" data sets; this may or may not be correct but I haven't seen any information on it). Assume that the robot frame is oriented so that the x axis points forwards.

  The "laser-angles" line tells you how many ranges will be on each line, the angle of the first, the angle of the last, and the angular step between each.

  It would probably be fine if you ignored the 181st laser datum in the "stanford" data set and shifted them by half a degree, but I've left it like this for completeness.

This script by default will produce the standard data file, but it can also produce an "odometry" file and a "laser" file which are suitable for plotting to see what the raw data look like. I plotted these files in gnuplot with the line:

plot "aces-laser.dat" using 1:2 with dots, "aces-odometry.dat" using 1:2 with lines

Here are the data files in our standard format:

• aces.zip
• stanford.zip
• intel.zip

FYI, here's the output my script produced when creating these files:

$ ./convert.rb aces3_publicb.rtl all
Processed 21957 lines which contained:
7375 odometry measurements
7375 laser scans (only 7374 used).

$ ./convert.rb intel.script all
Processed 40546 lines which contained:
26915 odometry measurements
13631 laser scans (only 13631 used).

$ ./convert.rb stanford-gates1.log all
Processed 179428 lines which contained:
20724 odometry measurements
118312 laser scans (only 20724 used).

If you run this script on your machine, you may need to change the first line so it has the path to ruby on your system. (For example, in the CS SUNOS environment, it is at /usr/local/bin/ruby.) You can also call it by running ruby directly, for example:

$ ruby convert.rb stanford-gates1.log all