Here is the information for my PhD presentation

Here is the video of my thesis rehearsal.  The audio pickup was not that good, so the sound is quiet.  Also the last minute is missing.

thesis.pdf

powerpoint.ppt

The videos are in the same order as in the presentation.

1emplanemapping.avi

2emnoisehires.avi

3pathcorrection.avi

4fullnonoiseperfect.avi

5fullnoise.avi

6pointfull.wmv

7dynamictestekfslam.avi

8dyanmicpointrlt.avi

9rltpoint4window2.avi

 

Here is a partial Victoria Park from iteration 1181 to 1642, basically where it is possible to link groups together since there is at least 3-4 points.  The average run time is 0.000296 seconds.  Multiple this by the total iterations of 7249 = 2.15 seconds to run the entire thing if the average stays consistent.

The segment of iterations relates to approximately 15-20 seconds of this video

Here are the results of the first run compared to the results of http://webdiis.unizar.es/~ppinies/research.htm.  The Relative Algorithm's result shown in blue squares are manually aligned to result from the above link.  It is likely the alignment is not perfect, but it shows that the Relative Algorithm results are generally consistent with the results from the EKF.  

.Here is another segment from iteration 4220-4634.  If your wondering how these intervals were picked, I took a look at intervals that had a minimum of 4 points and choose it that way.  The interval text file is here.  The average time was 0.00057 seconds due to using the 3 level grouping with dynamic detection.  There were mapping errors when only using the one level grouping that was used in the video above.  When using a 2 level grouping with dynamic detection the time went to 0.00040 seconds.  

Here are the results of the second run.