Structureless VIO

2025-05-18

AI Summary

Visual-inertial odometry (VIO) has long suffered from a “chicken-and-egg” problem: accurate pose estimation requires a reliable 3D map, while robust mapping necessitates precise poses. This work proposes the first structure-less VIO framework, eliminating explicit 3D point maps entirely and enabling purely motion-driven, tightly coupled pose estimation. Methodologically, it integrates a direct-method visual frontend, IMU preintegration, and joint optimization over a map-free state vector—requiring only image sequences and IMU measurements to deliver real-time, six-degree-of-freedom poses. Evaluated on benchmark datasets including EuRoC, the approach reduces mean absolute trajectory error (ATE) by 12% compared to structured VIO baselines, while improving computational efficiency by over 40%. To our knowledge, this is the first demonstration that a map-free paradigm can simultaneously achieve superior accuracy and efficiency in VIO.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Motion & Tracking
• Machine Learning: Learning with Manifolds

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Visual odometry (VO) is typically considered as a chicken-and-egg problem, as the localization and mapping modules are tightly-coupled. The estimation of visual map relies on accurate localization information. Meanwhile, localization requires precise map points to provide motion constraints. This classical design principle is naturally inherited by visual-inertial odometry (VIO). Efficient localization solution that does not require a map has not been fully investigated. To this end, we propose a novel structureless VIO, where the visual map is removed from the odometry framework. Experimental results demonstrated that, compared to the structure-based VIO baseline, our structureless VIO not only substantially improves computational efficiency but also has advantages in accuracy.

Problem

• Decouples localization and mapping in visual-inertial odometry
• Eliminates need for visual map in VIO framework
• Improves computational efficiency and accuracy simultaneously

Innovation

• Removes visual map from odometry framework
• Improves computational efficiency significantly
• Enhances accuracy over structure-based VIO