LBDU-VIO: Learned Bias Dynamics and Uncertainty for Visual-Inertial Odometry with Unreliable Vision

2026-09-30

AI Summary

This study addresses the limited robustness of traditional visual-inertial odometry (VIO) in visually degraded scenarios, which stems from random-walk bias assumptions and fixed noise parameters. We propose a learning-enhanced Multi-State Constraint Kalman Filter (MSCKF) framework that introduces neural ordinary differential equations (ODEs) for continuous-time IMU bias state propagation, replacing the conventional random-walk model, while enabling motion-adaptive measurement noise covariance prediction. The framework is trained end-to-end with unsupervised pose supervision. Evaluations on the EuRoC and TUM-VI benchmarks demonstrate that our method significantly outperforms mainstream baselines. Notably, during 10-second visual interruptions, it reduces relative position error by 25.1% compared to S-MSCKF, effectively enhancing localization accuracy under extreme degradation conditions.

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.

Problem

• Visual-Inertial Odometry
• Unreliable Vision
• IMU Bias Dynamics
• Uncertainty Modeling
• MSCKF

Innovation

• Visual-Inertial Odometry
• Neural ODE
• MSCKF
• Bias Dynamics
• Uncertainty Model