Investigation of ArUco Marker Placement for Planar Indoor Localization

TL;DR

This paper addresses planar indoor localization using ArUco fiducial markers with a monocular camera, systematically evaluating how marker placement—specifically the number, orientation, and distance to the camera—influences localization accuracy. It combines classic camera pose estimation via PnP with RANSAC and an offline intrinsic calibration, and introduces a lightweight adaptive Kalman filter that updates the measurement noise covariance $R^m$ based on the number of detected markers for real-time tracking. Empirical results show that placing 5 or more markers visible from a position significantly reduces worst-case errors, that markers farther from the image center and with certain orientations improve accuracy, and that the adaptive Kalman filter provides robust, smooth trajectories across varying visibility. The work provides practical guidance for marker deployment in indoor environments and highlights the trade-offs between marker quantity, placement geometry, and real-time tracking performance, with potential extensions to more complex setups and fusion with odometry.

Abstract

Indoor localization of autonomous mobile robots (AMRs) can be realized with fiducial markers. Such systems require only a simple, monocular camera as sensor and fiducial markers as passive, identifiable position references that can be printed on a piece of paper and distributed in the area of interest. Thus, fiducial marker systems can be scaled to large areas with a minor increase in system complexity and cost. We investigate the localization behavior of the fiducial marker framework ArUco w.r.t. the placement of the markers including the number of markers, their orientation w.r.t. the camera, and the camera-marker distance. In addition, we propose a simple Kalman filter with adaptive measurement noise variances for real-time AMR tracking.

Figures (6)

• Figure 1: In a) is our marker placement for marker placement evaluation, where markers form two circles (inner: markers 0-7, outer: markers 8-23) on the ground. Three sets of marker rotations (w.r.t. the closest chessboard lines in angle) are defined, where markers with IDs from $\{0, 8, 2, 12, 4, 16, 6, 20\}$ belong to the $90^\circ$ rotations, $\{1, 10, 3, 14, 5, 18, 7,22\}$ belong to the $45^\circ$ rotations, and $\{9, 11, 13, 15, 17, 19, 21, 23\}$ belong to the $22.5^\circ$ rotations. For real image collection, the camera is installed at a height of 1.334m or 2.991m (as in b)). The image collected at 1.334m misses a few markers due to limited camera .
• Figure 2: Marker placement (on ceiling) and rectangular movement (black arrows) for the tracking experiments.
• Figure 3: Error distribution for the circular marker experiments in configuration I with camera heights of a) 1.334m and b) 2.991m. Fiducial markers are encoded with the definitions from Fig. \ref{['fig:circles']}.
• Figure 4: Violin plot of the errors in $x$ for different numbers of markers and a camera height of 2.991m.
• Figure 5: Mean errors and confidence ellipses for a) all different numbers of markers and b) zoom into region with 4+ markers. The camera is installed at a height of 2.991m.