Replacing raw IMU inputs with level-crossing events on SE(3) pre-integrations improves generalization and accuracy of neural displacement priors for inertial odometry.
Rio: Rotation-equivariance supervised learning of robust inertial odometry
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Neural Inertial Odometry from Lie Events
Replacing raw IMU inputs with level-crossing events on SE(3) pre-integrations improves generalization and accuracy of neural displacement priors for inertial odometry.