A Cost Effective Indoor Positioning System using Wi Fi with Machine Learning on a PaaS Backend

Abstract

Indoor positioning systems (IPS) have become crucial in a variety of fields, including smart buildings, healthcare, and retail. However, many existing systems have high deployment costs, low energy efficiency, and restricted scalability. This work describes an inexpensive IPS that uses Wi-Fi received signal strength indicator (RSSI) data and machine learning to determine interior positions in real time. The system collects RSSI fingerprints using a lightweight Flutter-based mobile app (implemented with Flutter for cross-platform compatibility) before offloading computation-intensive activities to a Platform-as-a-Service (PaaS) backend developed with FastAPI and MongoDB. K-Nearest Neighbours (KNN) is the principal localisation algorithm, chosen because of its simplicity, versatility, and competitive accuracy. Experiments conducted indoors show encouraging results in terms of localisation accuracy, energy consumption, and cost scalability. This method demonstrates that accurate indoor location is possible without specialised hardware or expensive infrastructure, making it appropriate for large-scale, low-cost deployments.