Technology Category
Intelligent Robots: Localization, Mapping, and NavigationComputer Vision: Generative Adversarial Networks (GANs) for VisionNatural Language Processing: Generation
LiGen: GAN-Augmented Spectral Fingerprinting for Indoor Positioning
🤖 AI Summary
To address the poor robustness and low accuracy of Wi-Fi-based indoor localization under dynamic environmental conditions, this paper proposes an infrastructure-free spectral fingerprinting localization method. We pioneer the use of visible-light spectral intensity patterns as location fingerprints and innovatively introduce generative adversarial networks (GANs) to alleviate the small-sample bottleneck: we design two novel generative models—PointGAN for single-point data augmentation and FreeGAN for trajectory-level augmentation under free-moving scenarios—and integrate them with a multilayer perceptron (MLP) to form an end-to-end localization framework. Experiments demonstrate that the method achieves sub-meter average localization error (<0.8 m) in complex indoor environments, outperforming state-of-the-art Wi-Fi baselines by over 50%. Moreover, it exhibits strong robustness against illumination variations, occlusions, and other environmental disturbances, significantly expanding the practical applicability of tag-free, infrastructure-free indoor localization.
Technology Category
Application Category
📝 Abstract
Accurate and robust indoor localization is critical for smart building applications, yet existing Wi-Fi-based systems are often vulnerable to environmental conditions. This work presents a novel indoor localization system, called LiGen, that leverages the spectral intensity patterns of ambient light as fingerprints, offering a more stable and infrastructure-free alternative to radio signals. To address the limited spectral data, we design a data augmentation framework based on generative adversarial networks (GANs), featuring two variants: PointGAN, which generates fingerprints conditioned on coordinates, and FreeGAN, which uses a weak localization model to label unconditioned samples. Our positioning model, leveraging a Multi-Layer Perceptron (MLP) architecture to train on synthesized data, achieves submeter-level accuracy, outperforming Wi-Fi-based baselines by over 50%. LiGen also demonstrates strong robustness in cluttered environments. To the best of our knowledge, this is the first system to combine spectral fingerprints with GAN-based data augmentation for indoor localization.
Problem
Research questions and friction points this paper is trying to address.
Improves indoor localization accuracy using spectral fingerprints
Addresses limited spectral data with GAN-based augmentation
Outperforms Wi-Fi systems in stability and robustness
Innovation
Methods, ideas, or system contributions that make the work stand out.
Uses ambient light spectral fingerprints for localization
Employs GANs for data augmentation in positioning
Achieves submeter accuracy with MLP on synthetic data
🔎 Similar Papers
No similar papers found.
