From Global to Local: Cluster-Aware Learning for Wi-Fi Fingerprinting Indoor Localisation

TL;DR

This paper tackles indoor localisation by structuring Wi-Fi fingerprint data with clustering to enable local, coherent models. It introduces two clustering viewpoints (XYZ and RSSI) and two deployment strategies (building- and floor-level), plus a cluster-estimation mechanism based on the strongest APs. Across three public datasets and multiple ML models, building-based clustering with RSSI or XYZ features generally reduces 2D positioning error, though floor-detection accuracy can drop relative to baselines. The approach demonstrates scalable, flexible improvements in localisation and provides actionable guidance on hyperparameter choices and computational trade-offs for real-world deployment.

Abstract

Wi-Fi fingerprinting remains one of the most practical solutions for indoor positioning, however, its performance is often limited by the size and heterogeneity of fingerprint datasets, strong Received Signal Strength Indicator variability, and the ambiguity introduced in large and multi-floor environments. These factors significantly degrade localisation accuracy, particularly when global models are applied without considering structural constraints. This paper introduces a clustering-based method that structures the fingerprint dataset prior to localisation. Fingerprints are grouped using either spatial or radio features, and clustering can be applied at the building or floor level. In the localisation phase, a clustering estimation procedure based on the strongest access points assigns unseen fingerprints to the most relevant cluster. Localisation is then performed only within the selected clusters, allowing learning models to operate on reduced and more coherent subsets of data. The effectiveness of the method is evaluated on three public datasets and several machine learning models. Results show a consistent reduction in localisation errors, particularly under building-level strategies, but at the cost of reducing the floor detection accuracy. These results demonstrate that explicitly structuring datasets through clustering is an effective and flexible approach for scalable indoor positioning.

Figures (9)

• Figure 1: Overall architecture of the proposed clustering-based indoor positioning method, illustrating the training and localisation phases.
• Figure 2: Comparison of both approaches for creating the clusters in the 2nd floor of UJIIndoorLoc.
• Figure 3: Experimental setup employed to evaluate the localisation and floor detection performance of the proposed method.
• Figure 4: 2D positioning error percentage change of clustering approaches compared with the baseline approach in each dataset. A negative increment indicates that the positioning error has been improved compared with the baseline. Top figures are for the building strategy and bottom figures for floor strategy. Blue and red bars represent percentual change of XYZ clustering vs. Baseline and RSSI clustering vs Baseline, respectively.
• Figure 5: Floor detection rate percentage change of clustering approaches compared with the baseline approach in each dataset. A positive increment indicates that the FDR has been improved compared with the baseline. Top figures are for the building strategy and bottom figures for floor strategy. Blue and red bars represent percentual change of XYZ clustering vs. Baseline and RSSI clustering vs Baseline, respectively.