Visual-Inertial Odometry

Also known as VIO

Overview

Visual-inertial odometry (VIO) estimates a moving camera’s position and orientation by combining visual constraints with high-rate inertial measurements, enabling metric navigation when GPS is unavailable. Systems commonly use filtering methods such as MSCKF and error-state Kalman filters, or sliding-window and factor-graph optimization with IMU preintegration, feature tracking, and careful initialization. Newer approaches add learned feature tracking or depth, uncertainty-aware weighting, and continuous-time fusion; they are evaluated on benchmarks such as EuRoC, TUM-VI, and KITTI, as well as in specialized environments. A central challenge is improving robustness to poor lighting, dynamic scenes, sensor outages, and degenerate motion while maintaining observability, estimator consistency, and real-time operation on embedded hardware.

Latest papers

CardsList

64 papers8 papers in the last four weeks, down 38% on the four weeks before. 0.1% of all new papers.

  1. MDE-VIO: Enhancing Visual-Inertial Odometry Using Learned Depth Priors

    Feb 11, 2026Arda Alniak, Sinan Kalkan, Mustafa Mert Ankarali +2Inertial OdometryVisual-Inertial Odometry

  2. Onboard UAV State Estimation for Fast Flights Without External Localization Systems: A2RL Drone Racing Finalist Approach

    Feb 2, 2026Filip Novák, Matěj Petrlík, Matej Novosad +3Robot State EstimationVisual-Inertial Odometry

  3. DefVINS: Visual-Inertial Odometry for Deformable Scenes

    Jan 2, 2026Samuel Cerezo, Javier CiveraInertial OdometryVisual-Inertial Odometry

  4. SUPER - Sensitivity-based Uncertainty-aware Performance and Risk Assessment in Visual Inertial Odometry

    Dec 16, 2025Johannes A. Gaus, Daniel Häufle, Woo-Jeong BaekVisual-Inertial Odometry