search

surberj / Vision

15 stars 8 forks source link

ROVIO on 20160202 ticino datasets #41

Closed surberj closed 8 years ago

surberj commented 8 years ago

Without fine tuning of ROVIO all estimates immediately diverge during transition. By tuning initialization distance, covariances, number of features, patch levels, etc ROVIO can handle dataset sess001 but is drifting heavily:

sess001 // 131 degree lense // rovio.info saved // bag run with -s 122 sess001_pos

This is a very early result but increases hope that we can gain something out of the data :-).

surberj commented 8 years ago

The picture looks similar for the 72deg lense:

sess004 // 72 degree lense // bag run with -s 315 sess004_pos

ROVIO is not diverging but drifting heavily.

surberj commented 8 years ago

Comparison with PX4 EKF: see https://github.com/wingtra/vision/issues/96

surberj commented 8 years ago

OKVIS results by now also available (see wingtra/vision#96).