[Paper Review] SenseFlow: An Experimental Study for Tracking People

Kai Li, Chau Yuen+5 more|arXiv (Cornell University)|Jun 12, 2016

Mobile Crowdsensing and Crowdsourcing14 references3 citations

TL;DR

SenseFlow proposes a lightweight, passive people tracking system using probe requests from smartphones to monitor crowd density and track movement flows in real-world environments. By leveraging probe request intervals and RSSI-based localization, it reduces network overhead and improves tracking accuracy, achieving up to 92.9% tracking accuracy under optimal conditions (Android, screen on, Wi-Fi non-registered).

ABSTRACT

The main challenges in large-scale people tracking are the recognition of people density in a specific area and tracking the people flow path. To address these challenges, we present SenseFlow, a lightweight people tracking system. SenseFlow utilises off-the-shelf devices which sniff probe requests periodically polled by user's smartphones in a passive manner. We demonstrate the feasibility of SenseFlow by building a proof-of-concept prototype and undertaking extensive evaluations in real-world settings. We deploy the system in one laboratory to study office hours of researchers, a crowded public area in city to evaluate the scalability and performance "in the wild", and four classrooms in the university to monitor the number of students. We also evaluate SenseFlow with varying walking speeds and different models of smartphones to investigate the people flow tracking performance.

Motivation & Objective

• To develop a lightweight, scalable people tracking system using passive probe request sniffing from smartphones.
• To address network overload from high-volume probe request transmission in dense environments.
• To improve tracking accuracy by resolving the probe request overhearing problem where multiple gateways detect the same device.
• To evaluate the impact of smartphone operational modes (e.g., screen on/off, Wi-Fi registered/non-registered) on tracking performance.
• To enable real-time monitoring of people density and trajectory flow in diverse real-world settings such as labs, public areas, and classrooms.

Proposed method

• Deploy time-synchronized wireless gateway nodes (GNs) to passively collect probe requests from smartphones across different locations.
• Implement a Probe Request Interval-based Data Collection Scheme (PRI-DCS) that transmits only MAC address and timestamp, reducing network traffic.
• Use Received Signal Strength Indicator (RSSI) to resolve probe request overhearing by selecting the GN with the strongest signal as the device's location.
• Model user trajectories as sequences of GN IDs and apply the Longest Common Subsequence (LCS) algorithm to recognize and track movement paths.
• Aggregate data from all GNs on a central server to compute real-time people density and trajectory patterns.
• Use controlled and real-world experiments to validate the system under varying walking speeds, smartphone models, and operational modes.

Experimental results

Research questions

• RQ1 How does the probe request transmission interval vary across different smartphone models and operational modes, and how does this affect tracking accuracy?
• RQ2 To what extent does probe request overhearing degrade tracking performance, and how can it be mitigated using RSSI-based selection?
• RQ3 What is the impact of walking speed on the detection and tracking of individuals using probe requests?
• RQ4 How accurately can SenseFlow recognize predefined movement trajectories in real-world environments?
• RQ5 What is the performance of the PRI-DCS scheme in reducing network overhead while maintaining tracking fidelity?

Key findings

• SenseFlow achieves a maximum tracking accuracy of 92.9% for Android smartphones in the (screen on, Wi-Fi non-registered) mode, where probe requests are transmitted most frequently.
• In the (screen off, Wi-Fi registered) mode, tracking accuracy drops significantly to 44.3% for Android and 27.2% for iOS, due to increased probe request intervals.
• The (screen off, Wi-Fi non-registered) mode improves tracking accuracy for iOS and Windows phones compared to (screen on, Wi-Fi registered), as devices scan more frequently for networks.
• The successful trajectory recognition rate reaches 100% in the (screen on, Wi-Fi non-registered) mode, but drops to 60% in (screen on, Wi-Fi registered) and only 20% in (screen off, Wi-Fi registered) modes.
• The system effectively mitigates probe request overhearing by using RSSI to select the most likely gateway, improving localization accuracy.
• The PRI-DCS scheme significantly reduces network traffic by transmitting only essential data (MAC address and timestamp), enabling scalable deployment in large-scale environments.