Drift on ocotomap and robot wheel odometry

edit

octomap is a representation of input (3d data with transformations between them). It is the responsibility of the user to provide accurate transformations between two point clouds to create accurate 3D octomap based representations. Definitely one has to use some kind of localizations (amcl in 2D) and RGB-D SLAM (for example in 3D) to provide accurate transformations between two point clouds. Make sure that the algorithms perform loop closure, or else it may not give good 3D representations.

All type of odometry (either wheel or visual) accumulate errors over time and hence places need to be revisited, detected and accumulated error need to be corrected. This is generally done in a graph optimization framework after detecting the loop closures.

Good luck