[Paper Review] Fast and Reliable WiFi Fingerprint Collection for Indoor Localization

Fuqiang Gu, Milad Ramezani+4 more|arXiv (Cornell University)|Aug 1, 2020

Indoor and Outdoor Localization Technologies17 references4 citations

TL;DR

This paper proposes a fast, automated WiFi fingerprint collection method using smartphone sensors and a landmark graph to estimate reference point (RP) locations without user intervention. By leveraging inertial sensors and landmark-based PDR for location estimation, the method reduces site survey time by an order of magnitude and achieves a mean localization error of 1.5 meters—outperforming both manual and state-of-the-art Zee methods—while ensuring high accuracy through a belief-based quality metric.

ABSTRACT

Fingerprinting is a popular indoor localization technique since it can utilize existing infrastructures (e.g., access points). However, its site survey process is a labor-intensive and time-consuming task, which limits the application of such systems in practice. In this paper, motivated by the availability of advanced sensing capabilities in smartphones, we propose a fast and reliable fingerprint collection method to reduce the time and labor required for site survey. The proposed method uses a landmark graph-based method to automatically associate the collected fingerprints, which does not require active user participation. We will show that besides fast fingerprint data collection, the proposed method results in accurate location estimate compared to the state-of-the-art methods. Experimental results show that the proposed method is an order of magnitude faster than the manual fingerprint collection method, and using the radio map generated by our method achieves a much better accuracy compared to the existing methods.

Motivation & Objective

• To reduce the time and labor required for WiFi fingerprint site surveys, which are traditionally manual and error-prone.
• To eliminate the need for active user participation in fingerprint association, overcoming limitations of active crowdsourcing methods.
• To improve localization accuracy by using landmark graph-based PDR for automatic RP location estimation.
• To ensure reliability of the radio map by introducing a belief metric to filter low-confidence location estimates.
• To develop a low-cost, scalable solution suitable for deployment on resource-limited smartphones.

Proposed method

• Uses smartphone inertial sensors (accelerometer, gyroscope, magnetometer, barometer) to estimate step length and heading in real time.
• Employs a landmark graph-based PDR method to bound cumulative dead reckoning errors by calibrating with detected landmarks.
• Automatically associates collected WiFi fingerprints with estimated RP locations without requiring user input or GPS.
• Introduces a belief metric to evaluate the reliability of each location estimate, retaining only high-confidence RPs for radio map construction.
• Constructs a radio map using positive fingerprint representation: RSS values above a threshold τ are mapped to RSS_i - min, otherwise zero.
• Applies 1-NN matching with Euclidean and Sorensen distance metrics to compare fingerprints during online localization.

Experimental results

Research questions

• RQ1 Can a landmark graph-based method reduce the time and labor of WiFi fingerprint site surveys compared to manual methods?
• RQ2 Can automated fingerprint association using smartphone sensors achieve higher localization accuracy than existing methods?
• RQ3 How does the belief metric improve the reliability of estimated reference point locations in the radio map?
• RQ4 What is the impact of RSS threshold selection on localization performance in the proposed system?
• RQ5 Can the proposed method outperform state-of-the-art systems like Zee in both speed and accuracy?

Key findings

• The proposed method reduces fingerprint collection time by an order of magnitude—approximately 10 minutes for setup and data collection—compared to 90 minutes for manual methods.
• The mean localization error using the radio map generated by the proposed method is 1.5 meters, outperforming both the manual method (2.2 meters) and the Zee system (2.3 meters).
• The method achieves floor recognition accuracy comparable to Zee and significantly higher than manual methods with 5 or 10 scans per location.
• The optimal RSS threshold for minimizing localization error is -80 dBm, as values above or below this threshold increase error due to noise or missing APs.
• The belief metric effectively filters low-confidence location estimates, contributing to a more reliable and accurate radio map.
• The system is most effective in environments with sufficient spatial constraints; performance degrades in large, open indoor areas lacking landmarks.