analiza aiqbench
This research introduces LBDU-VIO, a learning-augmented framework that replaces standard random walk assumptions in Visual-Inertial Odometry with neural ODEs to model continuous bias dynamics and adaptive IMU uncertainty. By training these models on pose supervision without direct labels, the system significantly improves robustness during extended visual outages, demonstrating a 25.1% reduction in position error on EuRoC benchmarks. The work represents a meaningful advancement in integrating deep learning directly into classical state estimation pipelines for autonomous aerial systems.
LBDU-VIO: Learned Bias Dynamics and Uncertainty for Visual-Inertial Odometry with Unreliable Vision
arXiv:2609.39125v1 Announce Type: new Abstract: Visual-inertial odometry (VIO) for aerial robots relies on high rate inertial measurement unit (IMU) propagation between visual updates. However, conventional multi state constraint Kalman filters (MSCKFs) use random walk bias assumptions and fixed noise parameters, which can limit robustness when visual information is unreliable. To address this problem, we propose LBDU-VIO, a learning-augmented MSCKF with learned continuous time bias dynamics and an IMU uncertainty model. A neural ordinary differential equation (ODE) models continuous time bias dynamics to propagate the filter's bias states, replacing their random walk model. The IMU uncertainty model predicts motion adaptive measurement noise covariances for covariance propagation. Both models are trained with pose supervision without direct labels. Experiments on real world EuRoC and TUM-VI benchmarks show lower errors than representative visual-inertial baselines, including a 25.1% reduction in mean relative position error compared with S-MSCKF on EuRoC sequences with 10s visual outage.
Abstract:Visual-inertial odometry (VIO) for aerial robots relies on high rate inertial measurement unit (IMU) propagation between visual updates. However, conventional multi state constraint Kalman filters (MSCKFs) use random walk bias assumptions and fixed noise parameters, which can limit robustness when visual information is unreliable. To address this problem, we propose LBDU-VIO, a learning-augmented MSCKF with learned continuous time bias dynamics and an IMU uncertainty model. A neural ordinary differential equation (ODE) models continuous time bias dynamics to propagate the filter's bias states, replacing their random walk model. The IMU uncertainty model predicts motion adaptive measurement noise covariances for covariance propagation. Both models are trained with pose supervision without direct labels. Experiments on real world EuRoC and TUM-VI benchmarks show lower errors than representative visual-inertial baselines, including a 25.1% reduction in mean relative position error compared with S-MSCKF on EuRoC sequences with 10s visual outage.
Robotics (cs.RO)
arXiv:2609.39125 [cs.RO]
(or arXiv:2609.39125v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2609.39125 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 237 cuvinte
Subiecte conexe
\textbf{PLATO}: \emph{Preintegration Learning from Accurate Trajectory Observations} for Neural Inertial Odometry
arXiv · 9 sept. 2026
Does Robust VIO Need More Learning? Geometry-Verified Visual Measurements under Distribution Shift
arXiv · 21 iul. 2026
Uncertainty-Aware Adaptive Sensor Fusion for Autonomous Navigation
arXiv · 5 iun. 2026
Benchmarking Visual-Inertial Odometry in Subterranean Environments Under Sensor Degradation, Miscalibration, and Dynamic Occlusion
arXiv · 17 sept. 2026