[Paper Review] Autonomous WiFi Fingerprinting for Indoor Localization

Shilong Dai, Liang He+1 more|arXiv (Cornell University)|Nov 26, 2019

Indoor and Outdoor Localization Technologies25 references4 citations

TL;DR

AuF is an autonomous WiFi fingerprinting system that constructs indoor fingerprint databases without robot sojourns, using 2.4/5GHz signal correlation and spatial modeling to recover lost and abnormal signals. It reduces survey time and energy by 61–71% and 64–64% respectively, while maintaining baseline localization accuracy.

ABSTRACT

WiFi-based indoor localization has received extensive attentions from both academia and industry. However, the overhead of constructing and maintaining the WiFi fingerprint map remains a bottleneck for the wide-deployment of WiFi-based indoor localization systems. Recently, robots are adopted as the professional surveyor to fingerprint the environment autonomously. But the time and energy cost still limit the coverage of the robot surveyor, thus reduce its scalability. To fill this need, we design an AutonomousWiFi Fingerprinting system, called AuF, which autonomously constructs the fingerprint database with time and energy efficiency. AuF first conduct an automatic initialization process in the target indoor environment, then constructs the WiFi fingerprint database of in two steps: (i) surveying the site without sojourn, (ii) recovering unreliable signals in the database with two methods. We have implemented and evaluated AuF using a Pioneer 3-DX robot, on two sites of our $70$$ imes$$90$m$^2$ Department building with different structures and deployments of access points (APs). The results show AuF finishes the fingerprint database construction in 43/51 minutes, and consumes 60/82 Wh on the two floors respectively, which is a 64%/71% and 61%/64% reduction when compared to traditional site survey methods, without degrading the localization accuracy.

Motivation & Objective

• Address the high time and energy cost of traditional WiFi fingerprint map construction using human or robot surveyors.
• Overcome the limitations of robot surveying with sojourns, which increase time and power consumption due to frequent acceleration/deceleration.
• Enable fully autonomous, scalable deployment of WiFi fingerprinting in large indoor environments.
• Maintain high localization accuracy despite reduced signal collection during travel-without-sojourn surveys.
• Develop signal recovery techniques to compensate for unreliable measurements in a sojourn-free surveying regime.

Proposed method

• Conducts automatic floor recognition and map segmentation to enable efficient path planning for robot navigation.
• Performs site survey without stopping at reference points (travel-without-sojourn) to minimize time and energy consumption.
• Recovers lost 2.4/5GHz signals using strong empirical correlation between the two bands, leveraging dual-band WiFi access points.
• Detects abnormal WiFi samples using a spatial signal strength model that identifies outliers based on spatial consistency.
• Recovers abnormal signals via Gaussian process regression using previously constructed database data as prior knowledge.
• Applies region-wise signal recovery to limit computational load, ensuring linear scaling with survey area.

Experimental results

Research questions

• RQ1 Can a robot-based WiFi fingerprinting system achieve significant time and energy savings by eliminating sojourns during site surveys?
• RQ2 To what extent can 2.4GHz and 5GHz signal correlations be leveraged to recover lost signal measurements during high-speed surveys?
• RQ3 Can spatial modeling and database correlation effectively detect and recover abnormal WiFi signal samples in a sojourn-free setting?
• RQ4 Does the proposed signal recovery mechanism preserve localization accuracy compared to traditional fingerprinting methods?
• RQ5 Can the system scale efficiently to large indoor environments without incurring prohibitive computational costs?

Key findings

• AuF reduced site survey time by 64% on the 3rd floor and 71% on the 6th floor compared to traditional sojourn-based methods.
• Energy consumption was reduced by 61% on the 3rd floor and 64% on the 6th floor, with 60Wh and 82Wh consumed respectively.
• Localization accuracy using AuF’s fingerprint database showed no degradation: mean error was 2.4m (vs. 2.2m baseline) and max error 4.9m (vs. 7.5m baseline) with Bayesian inference.
• The 80th percentile error improved from 2.1m to 1.9m when using AuF’s data, indicating better overall performance distribution.
• Signal recovery computation took only 14s (3rd floor) and 23s (6th floor), which was negligible compared to survey time, validating scalability.
• The system achieved full autonomy through automatic floor recognition, path planning, and signal recovery without human intervention.