Publicação
Wi-Fi-based buildings occupancy detection, estimation, and location
| dc.contributor.author | Benites, Santiago Nogueira Mendes Mourão | |
| dc.contributor.institution | Faculty of Sciences | |
| dc.contributor.institution | Department of Informatics | |
| dc.contributor.supervisor | Cecílio, José Manuel da Silva | |
| dc.contributor.supervisor | Ferreira, Pedro Miguel Frazão Fernandes | |
| dc.date.accessioned | 2026-01-21T11:35:04Z | |
| dc.date.available | 2026-01-21T11:35:04Z | |
| dc.date.issued | 2025 | |
| dc.description | Tese de Mestrado, Segurança Informática, 2025, Universidade de Lisboa, Faculdade de Ciências | |
| dc.description.abstract | Smart buildings increasingly rely on accurate occupancy detection and indoor localisation to improve energy efficiency, enhance safety, and enable advanced services. Traditional Wi-Fibased fingerprinting approaches, while effective, often lack scalability and generalization, since models are typically tied to static environments and specific access point configurations. This thesis proposes a methodology for Wi-Fi-based building occupancy detection, estimation, and localisation that emphasises modularity, generalization, and security. The work introduces a geometric approach to Received Signal Strength Indicator (RSSI) fingerprinting, where physical spaces are divided into geometric subspaces and machine learning (ML) models are trained per subspace. By normalising RSSI values and abstracting distances, the method reduces dependency on specific access point placements, enabling models to be reused across different environments. Three experimental setups (indoor, garage, and outdoor) were used to evaluate the system, supported by extensive data collection with Raspberry Pi devices. Multiple ML models, including Random Forest (RF), Support Vector Machine (SVM), Decision Trees (DT), and Neural Networks (NNs), were tested, with Convolutional Neural Network (CNNs) further analysed for normalisation effects and feature integration. Results demonstrate that the proposed approach not only improves localisation accuracy but also enables crossdeployment applicability under certain constraints. Additionally, occupancy estimation experiments confirm the potential for integrating the methodology into real-world smart building systems. Despite limitations such as dataset size and environmental variability, the findings highlight the viability of geometric RSSI-based fingerprinting as a reusable and scalable solution for indoor positioning. The thesis concludes with recommendations for future work, including expanding datasets, exploring additional device and environment diversity, and refining continuous RSSI data collection methods. | en |
| dc.format | application/pdf | |
| dc.identifier.uri | http://hdl.handle.net/10400.5/116753 | |
| dc.language.iso | eng | |
| dc.subject | Wi-Fi localisation | |
| dc.subject | occupancy detection | |
| dc.subject | RSSI fingerprinting | |
| dc.subject | ML | |
| dc.subject | smart buildings | |
| dc.title | Wi-Fi-based buildings occupancy detection, estimation, and location | en |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess |
Ficheiros
Principais
1 - 1 de 1
A carregar...
- Nome:
- TM_Santiago_Benites.pdf
- Tamanho:
- 3.05 MB
- Formato:
- Adobe Portable Document Format