CrowdLOC-S: Crowdsourced Seamless Localization Framework Based on CNN-LSTM-MLP Enhanced Quality Indicator

Year: 2024

Venue: Expert Systems with Applications, Volume 243, Article 122852

DOI: 10.1016/j.eswa.2023.122852

Local Availability

No open-access full-text location was exposed through OpenAlex at retrieval time. Attempts to fetch the ScienceDirect PDF endpoint from the current environment were blocked on 2026-04-16, so this local page records the paper's metadata and summary.

Summary

This paper presents CrowdLOC-S, a seamless indoor/outdoor localization framework that combines crowdsourced Wi-Fi fingerprinting, GNSS, and low-cost MEMS sensors. Its design includes a data-and-model dual-driven trajectory estimator, a hybrid 1D-CNN, Bi-LSTM, and MLP quality indicator for crowdsourced trajectory evaluation and Wi-Fi database construction, transfer learning for scene switching and error prediction, and EKF-based multi-source fusion for unified meter-level positioning across complex urban environments.

Source Links