Crowdsensing-based WiFi Indoor Localization
Crowdsensing-based WiFi Indoor Localization
Embedded Files
Publications:
[1] [Best Paper Award] Barnwal S, and Peng W. Crowdsensing-based WiFi Indoor Localization using Feed-forward Multilayer Perceptron Regressor. In IEEE 2nd International Conference on Computational Intelligence in Data Science (ICCIDS), 2019.
Overview
In this research, we trained MLP Regressor on crowdsensed fingerprinting dataset using scikit-learn and localized cellphone users based on the RSS perceived from a thousand Access Points.
Contrary to popular literature, the work indicated that training a single deep neural network for all kinds of cellphones rather than separately trained models is optimal. This makes it easily deployable on devices having limited hardware/software capabilities.
Addressing device heterogeneity, our algorithm achieves 80% better positioning accuracy than the probabilistic localization algorithms when encountered with a new device/software.
Dataset
”Wi-Fi Crowdsourced Fingerprinting Dataset for Indoor Positioning.” [Link], comprises of 4648 fingerprints that is collected with 21 devices in a 5-storeyed university building. These signals were received by 992 fixed receivers(Access Points) located in the building.
Duration: May 2018 - Feb 2019
Status: Completed
Members: Simran Barnwal, Dr. Wei Peng
Note: The project was started during the summer break of 2018. This 12-weeks long internship at the University of Regina, Canada was funded by the very prestigious Mitacs Globalink Program. The research project was accepted at ICCIDS, 2019 and earned the Best Paper Award for its novelty.
Results: Comparison of mean 2D, mean 3D and Floor Detection rate(%) for our method vs state of the art methods.
Proposed Model: CDF of mean 3-D positioning error for test cases of each device
State of the art: Impact of the crowdsensing device type and the number of measurements per device on the positioning error *
Proposed Model: CDF of mean 3-D positioning error for all test cases irrespective of device
State of the art: CDF of mean 3-D positioning error for all test cases irrespective of device *
*Reference for state of art diagrams: Peng Z, Richter P, Leppäkoski H, Lohan E-S. ''Analysis of Crowdsensed WiFi Fingerprints for Indoor Localization.'' in Proceedings of the 21st Conference of Open Innovations Association FRUCT. Helsinki, Finland: FRUCT. 2017. p. 268-277.
Google Sites
Report abuse