I've got the Area Learning C API example running on my Google Tango but the pose after localization with respect to the ADF seems to be inaccurate: Below is an image of the outputted Tango position plotted alongside "ground truth" Tango position captured by motion capture sensors. I am supposing that this is because of accumulated drift, which is supposed to be minimal after localization.
The pose I am outputting is contained in the adf_T_device_pose_ variable which is updated with the pose data returned by the onPoseAvailable callbacks in area_learning_app.cc.
Does this variable contain drift-corrected pose data as soon as it is updated?
