Model-based Neural Data Augmentation for sub-wavelength Radio Localization

2025-06-05

AI Summary

In complex wireless environments dominated by non-line-of-sight (NLoS) propagation, conventional fingerprinting-based localization suffers from low accuracy, excessive memory overhead, and high computational complexity when leveraging machine learning methods. Method: This paper proposes a model-driven generative channel modeling framework. It introduces an interpretable, model-driven neural network as a lightweight generative channel model to enable both efficient fingerprint database augmentation and structured compression. The approach jointly integrates neural data augmentation with optimized fingerprint matching. Contribution/Results: The method achieves sub-wavelength localization accuracy—improving upon conventional fingerprinting by several orders of magnitude—while reducing fingerprint storage requirements by one order of magnitude. It effectively breaks the longstanding accuracy–memory trade-off bottleneck inherent in NLoS scenarios, offering a scalable, resource-efficient solution for high-precision indoor localization.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Machine Learning: Learning on the Edge & Model Compression
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Large-scale security measurements

Abstract

The increasing deployment of large antenna arrays at base stations has significantly improved the spatial resolution and localization accuracy of radio-localization methods. However, traditional signal processing techniques struggle in complex radio environments, particularly in scenarios dominated by non line of sight (NLoS) propagation paths, resulting in degraded localization accuracy. Recent developments in machine learning have facilitated the development of machine learning-assisted localization techniques, enhancing localization accuracy in complex radio environments. However, these methods often involve substantial computational complexity during both the training and inference phases. This work extends the well-established fingerprinting-based localization framework by simultaneously reducing its memory requirements and improving its accuracy. Specifically, a model-based neural network is used to learn the location-to-channel mapping, and then serves as a generative neural channel model. This generative model augments the fingerprinting comparison dictionary while reducing the memory requirements. The proposed method outperforms fingerprinting baselines by achieving sub-wavelength localization accuracy, even in NLoS environments. Remarkably, it offers an improvement by several orders of magnitude in localization accuracy, while simultaneously reducing memory requirements by an order of magnitude compared to classical fingerprinting methods.

Problem

• Improving radio localization accuracy in NLOS environments
• Reducing computational complexity in machine learning-assisted localization
• Enhancing fingerprinting-based localization with model-based neural networks

Innovation

• Model-based neural network for channel mapping
• Generative neural channel model augmentation
• Sub-wavelength accuracy with reduced memory