Technology Category
Intelligent Robots: Localization, Mapping, and NavigationMachine Learning: Hardware-aware MLComputer Vision: Motion & Tracking
Magnetic Indoor Localization through CNN Regression and Rotation Invariance
🤖 AI Summary
This work addresses the degradation in indoor localization accuracy caused by device orientation variations in magnetometer-based systems. To overcome this challenge, the authors propose MagNetS/XL, a lightweight dilated convolutional neural network that leverages rotation-invariant magnetic features—specifically, the magnetic field magnitude $M_n$ and its projection onto the gravity axis $M_g$—to directly regress position coordinates in an end-to-end manner. The approach eliminates the need for device pose alignment or additional infrastructure, relying solely on two-dimensional rotation-invariant inputs to achieve high robustness in real-world environments. Evaluated on the MagPie dataset, MagNetS attains state-of-the-art or comparable localization accuracy with only one-third of the model parameters, offering an efficient solution suitable for mobile deployment without compromising performance.
Technology Category
Application Category
📝 Abstract
Indoor positioning is an essential technology for a wide range of applications in GNSS-denied environments, including indoor navigation and IoT systems. Combining convolutional neural networks (CNNs) and magnetic field-based features offers a low-cost, infrastructure-free solution for precise positioning. While magnetic fingerprints are a promising approach for indoor positioning, models trained on raw 3D magnetometer data are highly sensitive to device orientation. We address this by using two rotation invariant features derived from the 3D magnetic field: the norm (Mn) and the projection onto the gravity axis (Mg). We train a lightweight 7-layer dilated CNN (MagNetS/XL) on magnetic sequences to directly regress (x, y) positions. Using the MagPie dataset (three buildings, handheld trajectories), we systematically evaluate fixed and random rotations of test and/or train data. Raw 3D inputs (Mx, My , Mz) exhibit isotropic error increases under fixed 90° rotations and further degrade with growing random rotations. In contrast, 2D (Mn, Mg) inputs maintain rotation invariant accuracy and surpass the 3D inputs once rotation exceeds building-specific thresholds for three reference buildings: 0° for Loomis (large), 5° for Talbot (medium), and 6° for CSL (small). MagNetXL achieves or exceeds state-of-the-art accuracy on the MagPie dataset, and MagNetS delivers similar performance with roughly one third of the parameters, favoring mobile deployment. These results show that the robustness gained from rotation invariant inputs outweighs the loss of input dimensionality in realistic usage, allowing mapping and localization without orientation alignment or added infrastructure.
Problem
Research questions and friction points this paper is trying to address.
indoor localization
magnetic field
rotation invariance
device orientation
magnetic fingerprinting
Innovation
Methods, ideas, or system contributions that make the work stand out.
rotation invariance
magnetic indoor localization
CNN regression
lightweight neural network
magnetic fingerprinting
🔎 Similar Papers
No similar papers found.
