[Paper Review] FedLoc: Federated Learning Framework for Data-Driven Cooperative Localization and Location Data Processing

Feng Yin, Zhidi Lin+6 more|arXiv (Cornell University)|Mar 8, 2020

Indoor and Outdoor Localization Technologies118 references14 citations

TL;DR

FedLoc proposes a federated learning framework for privacy-preserving, data-driven cooperative localization and location data processing, enabling mobile users to collaboratively train high-accuracy localization models without sharing raw location data. By leveraging distributed training with deep neural networks and Gaussian processes, FedLoc achieves near-centralized performance in localization and spatio-temporal prediction while preserving user privacy.

ABSTRACT

In this overview paper, data-driven learning model-based cooperative localization and location data processing are considered, in line with the emerging machine learning and big data methods. We first review (1) state-of-the-art algorithms in the context of federated learning, (2) two widely used learning models, namely the deep neural network model and the Gaussian process model, and (3) various distributed model hyper-parameter optimization schemes. Then, we demonstrate various practical use cases that are summarized from a mixture of standard, newly published, and unpublished works, which cover a broad range of location services, including collaborative static localization/fingerprinting, indoor target tracking, outdoor navigation using low-sampling GPS, and spatio-temporal wireless traffic data modeling and prediction. Experimental results show that near centralized data fitting- and prediction performance can be achieved by a set of collaborative mobile users running distributed algorithms. All the surveyed use cases fall under our newly proposed Federated Localization (FedLoc) framework, which targets on collaboratively building accurate location services without sacrificing user privacy, in particular, sensitive information related to their geographical trajectories. Future research directions are also discussed at the end of this paper.

Motivation & Objective

• To address the challenge of achieving high-precision localization in complex environments like indoor malls and urban areas using data-driven models.
• To enable collaborative model training across mobile users without compromising user privacy, especially regarding sensitive geographical trajectories.
• To overcome computational and communication bottlenecks in training large-scale localization models on decentralized, non-i.i.d. data.
• To extend federated learning to new applications in cooperative localization, target tracking, and spatio-temporal traffic prediction.
• To provide a unified framework that integrates distributed optimization, privacy-preserving techniques, and real-world deployment considerations for location services.

Proposed method

• Utilizes federated learning to train global models using local data collected by mobile users, avoiding raw data transmission.
• Employs deep neural networks (DNNs) and Gaussian processes (GPs) as core learning models for modeling complex localization and spatio-temporal patterns.
• Applies distributed optimization techniques with hyper-parameter tuning to handle non-i.i.d. data and unbalanced data sizes across users.
• Reduces communication overhead via model quantization (e.g., 8-bit precision) and signSGD to minimize gradient transmission.
• Incorporates robust optimization and estimation techniques to mitigate risks from malicious or unreliable user contributions.
• Uses data augmentation with canonical parametric models to improve performance when local data are scarce.

Experimental results

Research questions

• RQ1 Can federated learning achieve near-centralized localization accuracy while preserving user privacy in cooperative localization systems?
• RQ2 How can distributed learning models like DNNs and GPs be effectively applied to real-world location services such as indoor fingerprinting and outdoor GPS tracking?
• RQ3 What are the key challenges in communication efficiency, data heterogeneity, and model convergence in federated localization, and how can they be addressed?
• RQ4 How can federated learning be extended to handle multimodal data, such as combining wireless and visual measurements for improved localization robustness?
• RQ5 What role can social graph structures and graph-based models play in enhancing participation and performance in federated localization?

Key findings

• FedLoc achieves near-centralized performance in data fitting and prediction across multiple use cases, including indoor fingerprinting, outdoor GPS tracking, and wireless traffic prediction.
• Model quantization to 8-bit precision and signSGD reduce communication load by up to a factor of 32 without significant performance degradation.
• Robust distributed optimization techniques help maintain model accuracy even when some users contribute faulty or malicious updates.
• Data augmentation using canonical parametric models improves model generalization when local data are limited or sparse.
• The integration of wireless and visual measurements enhances localization accuracy and robustness, especially in challenging tracking scenarios.
• The framework is scalable and practical, supported by advances in 5G, edge AI chips, and high-precision maps, enabling real-world deployment.