Withdrawn arXiv Paper Proposed WiFi-Based System for Determining Building Floors Without Prior Layout Data

A research paper submitted to arXiv proposed an automated approach for distinguishing floor levels in indoor spaces using only WiFi signal data, without requiring advance knowledge of a building's layout or floor count.

The Problem#

Pinpointing a user's exact vertical position—particularly which floor they occupy—has long been a stubborn challenge for indoor navigation systems, especially in buildings where GPS signals cannot reach. While horizontal positioning within a floor has become reasonably reliable, determining elevation remains difficult when only WiFi signal strength measurements are available.

The Proposed Method#

The researchers framed the challenge as a graph problem. Each WiFi measurement point was treated as a node in a network, with connections drawn between measurements that showed similar signal strength patterns or appeared consecutively in user movement trails. The team then applied a technique called Node2Vec to learn representations of these nodes that captured their structural relationships, followed by automated clustering to group measurements by floor level.

Notably, the approach required no prior input about the building itself—no floor count, no architectural blueprints, no metadata—making it potentially portable across different structures.

Testing and Results#

The method was evaluated on publicly available indoor positioning datasets, including a newly published Huawei University Challenge 2021 benchmark and a modified version of the UJIIndoorLoc dataset. According to the abstract, the results showed the framework could successfully extract vertical structure information from signal-only data.

Current Status#

The paper has been withdrawn from arXiv as of October 2026, after multiple revisions since its May 2025 submission.