Estimating Occupancy in Public Spaces by Exploring WiFi Frames
Student
João Freitas (B.S.)
Abstract
This academic study explores the potential of WiFi frame analysis for estimating occupancy levels across diverse environments, from controlled settings to unregulated public spaces. Leveraging smartphones and connected devices, we analyze the frequency of IEEE 802.11 management frames to discern patterns of human presence over time.
The applications of this study include knowing the live occupation of space without being intrusive in terms of privacy, studying levels of occupation over multiple days, and allowing people to know if a certain space is crowded before reaching that place.
During the development of this thesis, a device was created capable of intercepting IEEE 802.11 traffic and transporting a copy of it to a server for further machine learning-based regression models to analyze.
Our research methodology includes case studies in uncontrolled environments, such as the Académica da Madeira main offices and a research hardware laboratory, assessing the effectiveness of various statistical and machine learning based regression models such as Linear Regression, Polynomial Regression, Multi-Layer Perceptrons, Random Forest, Gradient Boosting, Ada Boosting, Support Vector Machines, and K-Nearest Neighbors.
At the end of this study, the device was stable, and the machine learning models could estimate the occupancy levels of public spaces successfully and with a good degree of trust; the Gradient Boosting ended up being the machine learning model with the best performance for our problem and data.
More information
- Date
- 2024
