IMU Preintegration on Manifold for Efficient Visual-Inertial Maximum-a-Posteriori Estimation

Video accompanying the paper:
Christian Forster, Luca Carlone, Frank Dellaert, Davide Scaramuzza, "IMU Preintegration on Manifold for Efficient Visual-Inertial Maximum-a-Posteriori Estimation", Robotics: Science and Systems (RSS), Rome, 2015.
PDF: rpg.ifi.uzh.ch/docs/RSS15_Fors...
Supplementary Material: rpg.ifi.uzh.ch/docs/RSS15_Fors...
Abstract:
Recent results in monocular visual-inertial navigation (VIN) have shown that optimization-based approaches outperform filtering methods in terms of accuracy due to their capability to relinearize past states. However, the improvement comes at the cost of increased computational complexity. In this paper, we address this issue by preintegrating inertial measurements between selected keyframes. The preintegration allows us to accurately summarize hundreds of inertial measurements into a single relative motion constraint. Our first contribution is a preintegration theory that properly addresses the manifold structure of the rotation group and carefully deals with uncertainty propagation. The measurements are integrated in a local frame, which eliminates the need to repeat the integration when the linearization point changes while leaving the opportunity for belated bias corrections. The second contribution is to show that the preintegrated IMU model can be seamlessly integrated in a visual-inertial pipeline under the unifying framework of factor graphs. This enables the use of a structureless model for visual measurements, further accelerating the computation. The third contribution is an extensive evaluation of our monocular VIN pipeline: experimental results confirm that our system is very fast and demonstrates superior accuracy with respect to competitive state-of-the-art filtering and optimization algorithms, including off-the-shelf systems such as Google Tango.

[@NikolausDemmel]

Impressive work. Looking forward to Rome.

[@SiddharthJhakaas]

Impressive!

[@weikunzhen6386]

Very impressive. I have a question about the data association. How the landmark outliers are handled in the structureless formulation?

[@bergonius]

Can this algorithm be applied to stereo video feed? Would it work better?

[@3243234455422]

bergonius yes, it can be applied to stereo also. it works better because the scale is directly observable with two cameras.

[@mikejmills1]

Very interesting work. You said you released the code in GTSAM 4.0 but I can't find it. Is it in the bitbucket repo?

[@3243234455422]

Mike Mills yes, it's in gtsam/navigation/ImuFactor

[@mikejmills1]

Christian Forster Thank you

[@offmeds2nite]

Impressive. Is the proccessing occurring on the data stream happening on-board, or offline?

[@3243234455422]

***** we run the experiments off-line on a laptop computer. we are working on making it work on the drones...

[@alexandergrau887]

Did you succeed to get it working in realtime on the drones? (and if so, what hardware did you use and what FPS did you achieve)

[@NikolausDemmel]

One more question on the released implementaion. Which is the structureless vision factor referenced in the paper?

[@ailabRPG]

Nikolaus Demmel They are called SmartProjectionFactors in GTSAM :)

[@NikolausDemmel]

ailabRPG They are pretty smart ;-)
Thanks for the hint!

[@akunpang8625]

Is there any access to the staircase dataset ?

[@rushiqian8874]

Is the code open source?