City 1000

The first dataset that we will look at is the commonly used city10000. This is a simulated dataset, which was original pose graph can be seen below.

city10000 photo city10000_zps0oc5nogs.png

Pose Graph for City 10,000


Interestingly, the Levenberg–Marquardt least-squares would not work on this specific dataset. So, for this dataset, the Gauss–Newton implementation was used. A side-by-side comparison of the L-M and G-N algorithms attempting to optimize the city10000 graph is shown below ( Gauss-Newton implementation is shown on the right ).



Using the initial error free pose graph shown above, multiple false constraints were added to evaluate the performance of the robust optimization scheme. As a reference, the results for traditional $L^2$ optimization when 10 false constraints are present is $\mathcal{X}^2$ = 21634.

city10000Comp photo city10000Comp_zpsh2vu19tw.png

CSAIL

This dataset is built from raw data acquired at the MIT CSAIL building (the relative pose measurements are also available here )

man10000

Inital Pose Graph for CSAIL


Again, using the initial error free pose graph shown above, multiple false constraints were added to evaluate the performance of the robust optimization scheme. The results for DCS and switch factors are shown below.

csailComp.png photo csailComp_zpssqvk4aug.png



Freiburg Building

This pose graph is generated from raw data acquired at the Freiburg Building (the relative pose measurements are also available here )

man10000

Inital Pose Graph for FR079

Faults were added to the fault free graph in the same manner as state before. The results for DCS and switch factors are shown below.

fr079Comp photo fr079Comp_zpsgno9zeok.png



Freiburg Clinic

man10000

Inital Pose Graph for Freiburg Clinic



Next, I’m going to move to Max-Mix model