Impact of intialization on pose graph SLAM

[L D]

Hello,

in the example provided in gtsam online tutorial section 4, graph optimization requires initial estimate pose values (¤ 2.4 Non-linear Optimization in GTSAM).

Suppose we have detected a loop closure event and added this constraint to the graph. After optimization, all poses in the graph are updated. Do we have to replace the former initial pose estimates with the poses provided by the optimized graph ? Does the graph still rely on the former pose estimates ?

Regards.