Long-Term WiFi Fingerprinting Dataset for Research on Robust Indoor Positioning

TLDR

WiFi fingerprinting, a leading indoor positioning technique, suffers from limited robustness to short‑ and long‑term signal changes and from poor reproducibility of new methods. This paper introduces a WiFi RSS database designed to support research that tackles these robustness and reproducibility challenges. A professional collected consecutive fingerprints at fixed positions and orientations over 15 months, producing monthly training datasets and five types of test datasets to enable analysis of short‑ and long‑term signal variations. The database, along with supporting materials and software, is available on Zenodo under an open‑source MIT license, facilitating its use.

Abstract

WiFi fingerprinting, one of the most popular methods employed in indoor positioning, currently faces two major problems: lack of robustness to short and long time signal changes and difficult reproducibility of new methods presented in the relevant literature. This paper presents a WiFi RSS (Received Signal Strength) database created to foster and ease research works that address the above-mentioned two problems. A trained professional took several consecutive fingerprints while standing at specific positions and facing specific directions. The consecutive fingerprints may enable the study of short-term signals variations. The data collection spanned over 15 months, and, for each month, one type of training datasets and five types of test datasets were collected. The measurements of a dataset type (training or test) were taken at the same positions and directions every month, in order to enable the analysis of long-term signal variations. The database is provided with supporting materials and software, which give more information about the collection environment and eases the database utilization, respectively. The WiFi measurements and the supporting materials are available at the Zenodo repository under the open-source MIT license.