Pervasive Access Point Localization Utilizing Crowdsourced POI Data and Wi-Fi Scans

The Hong Kong University of Science and Technology
Department of Computer Science and Engineering


MPhil Thesis Defence


Title: "Pervasive Access Point Localization Utilizing Crowdsourced POI Data 
and Wi-Fi Scans"

By

Mr. Wing Ho CHEUNG


Abstract:

The locations of Wi-Fi access points (APs) are a valuable piece of 
information for applications such as indoor positioning, digital twin and 
router placement. Existing approaches to localize APs often require 
specialized devices, much manual data calibration, additional training, or 
prior knowledge of some AP locations in the venue, which greatly hampers 
their pervasive deployment. To overcome that, we propose LAPIS, a novel 
scalable approach that localizes APs using crowdsourced data of points of 
interest (POIs) and Wi-Fi scans. LAPIS first effectively matches SSIDs 
(Service Set Identifiers) of the Wi-Fi scans with the POI names using 
retrieval-augmented generation (RAG) on large language model (LLM), finding 
the first-order approximation of some APs. Given the results and by 
formulating a least-squares optimization problem, it subsequently localizes 
all the matched and unmatched APs. Demonstrating in leading malls in Hong 
Kong of area more than 90,000 m2, 6,640 APs and 6,000 crawled 
POIs, we show that LAPIS achieves an impressive POI matching precision of 
95.8%, and cuts AP localization error significantly by 45.7% and 56.1% as 
compared with two recent approaches, respectively.


Date:                   Tuesday, 20 January 2026

Time:                   4:30pm - 6:30pm

Venue:                  Room 3494
                        Lifts 25/26

Chairman:               Prof. Raymond WONG

Committee Members:      Prof. Gary CHAN (Supervisor)
                        Dr. Chaojian LI