Hierarchical Stage-Wise Training of Linked Deep Neural Networks for Multi-Building and Multi-Floor Indoor Localization Based on Wi-Fi RSSI Fingerprinting

TL;DR

This work tackles large-scale indoor localization across multiple buildings and floors using Wi‑Fi RSSI fingerprinting. It introduces a hierarchical stage-wise training (HST) framework that trains linked neural networks in stages, letting lower-level floor-location models leverage higher-level building/floor knowledge. The approach is demonstrated with both a DNN and a CNNLoc variant, achieving a 3D error of $8.19$ m and $8.71$ m respectively on the UJIIndoorLoc database, and showing improved performance over conventional training while maintaining practical training times. The results highlight the framework’s scalability, architecture-agnostic applicability, and potential to extend to other wireless sensing modalities, supporting accurate, on-device indoor localization in smart-city environments.

Abstract

In this paper, we present a new solution to the problem of large-scale multi-building and multi-floor indoor localization based on linked neural networks, where each neural network is dedicated to a sub-problem and trained under a hierarchical stage-wise training framework. When the measured data from sensors have a hierarchical representation as in multi-building and multi-floor indoor localization, it is important to exploit the hierarchical nature in data processing to provide a scalable solution. In this regard, the hierarchical stage-wise training framework extends the original stage-wise training framework to the case of multiple linked networks by training a lower-hierarchy network based on the prior knowledge gained from the training of higher-hierarchy networks. The experimental results with the publicly-available UJIIndoorLoc multi-building and multi-floor Wi-Fi RSSI fingerprint database demonstrate that the linked neural networks trained under the proposed hierarchical stage-wise training framework can achieve a three-dimensional localization error of 8.19 m, which, to the best of the authors' knowledge, is the most accurate result ever obtained for neural network-based models trained and evaluated with the full datasets of the UJIIndoorLoc database, and that, when applied to a model based on hierarchical convolutional neural networks, the proposed training framework can also significantly reduce the three-dimensional localization error from 11.78 m to 8.71 m.

Figures (5)

• Figure 1: A comparison of the location information of a Wi-Fi RSSI fingerprint between multi-building & multi-floor and single-building & single-floor indoor localization, the former of which shows a hierarchical structure of the location.
• Figure 2: DNN architectures for large-scale multi-building and multi-floor indoor localization, where the SDAE of rela_dnn_01 is replaced by SAE for simplicity.
• Figure 3: CNN architectures for large-scale multi-building and multi-floor indoor localization.
• Figure 4: An overview of the HST of the linked-DNN model for multi-building and multi-floor indoor localization.
• Figure 5: An overview of the HST of the linked-CNNLoc model for multi-building and multi-floor indoor localization.