PL-VIWO: A Lightweight and Robust Point-Line Monocular Visual Inertial Wheel Odometry

TL;DR

PL-VIWO addresses the challenge of robust localization for ground robots in outdoor environments by tightly coupling monocular visual-inertial data with wheel odometry and a novel line-feature pipeline. It introduces Motion Consistency Check (MCC) to filter dynamic points and a 2D line processing workflow that exploits point-line geometry for fast, robust line tracking and triangulation within an MSCKF framework. The approach demonstrates improved localization accuracy and efficiency over state-of-the-art methods on KAIST urban datasets, particularly in texture-poor and dynamic settings, while maintaining real-time performance on resource-constrained hardware. The work highlights the practical impact of incorporating well-integrated line features and MCC in monocular VIWO, with potential extensions to stereo and richer point-line constraints for further gains.

Abstract

This paper presents a novel tightly coupled Filter-based monocular visual-inertial-wheel odometry (VIWO) system for ground robots, designed to deliver accurate and robust localization in long-term complex outdoor navigation scenarios. As an external sensor, the camera enhances localization performance by introducing visual constraints. However, obtaining a sufficient number of effective visual features is often challenging, particularly in dynamic or low-texture environments. To address this issue, we incorporate the line features for additional geometric constraints. Unlike traditional approaches that treat point and line features independently, our method exploits the geometric relationships between points and lines in 2D images, enabling fast and robust line matching and triangulation. Additionally, we introduce Motion Consistency Check (MCC) to filter out potential dynamic points, ensuring the effectiveness of point feature updates. The proposed system was evaluated on publicly available datasets and benchmarked against state-of-the-art methods. Experimental results demonstrate superior performance in terms of accuracy, robustness, and efficiency. The source code is publicly available at: https://github.com/Happy-ZZX/PL-VIWO

Figures (8)

• Figure 1: PL-VIWO in KAIST Urban32. $\textbf{Left}$: Top-down view of the trajectory with line mapping results. $\textbf{Top-Right}$: Point tracking results, where red dots connected by lines represent observations of point features within the sliding window, and green indicates those failed in triangulation or MCC. $\textbf{Middle-Right}$: Zoomed-in trajectory (green) with lines (blue). $\textbf{Bottom-Right}$: Point-line pairing results, where circles with the same color line indicate point features that lie on the corresponding line feature.
• Figure 2: The framework of PL-VIWO system is divided into four components, each represented by different color-coded boxes. The boxes highlighted with bold red text indicate the key contributions of this work.
• Figure 3: 2D Line process results in KAIST Urban 28. (a) The red line segments represent the detection result. (b) The colored lines represent the classification results: red, green, and blue indicate lines parallel to the x-, y-, and z-axes of the IMU frame, respectively. Non-parallel lines are omitted from the diagram. The red circle around the centre of the image represents the vanishing point corresponding to the x-axis while y and z lie outside the image. (c) Circles lying on a line with the same color represent point features assigned with a line. (d) The colored lines in the figure are formed by connecting the midpoints of all observed line segments within the sliding window. The redder line means more recent.
• Figure 4: Line triangulation (a) Triangulation from planes. (b) Degenerate motion for line feature triangulation. Cameras stay in the same plane $\pi$. (c) Triangulation from points and direction.
• Figure 5: Trajectories of the KAIST Urban 31(PL-VINS failed).