Locating an Unknown Transmitter Using a Massive Radio Network : Determining Position Using Received Signal Strength

Sammanfattning

The increased use of wireless systems has created a complex electromagnetic environment, increasing the need for accessible methods to monitor and secure the radio frequency spectrum. While high precision localisation systems exist, they often require expensive hardware and specialised synchronisation that is neither widely available nor easily deployed. This thesis investigates a cost effective alternative using a distributed network of commercial off the shelf (COTS) sensors based solely on observing received signal strength (RSS). Central to this approach is the strategic selection of hardware and the implementation of a robust distributed system architecture, ensuring that low cost nodes can function reliably as a cohesive monitoring network. The problem is attacked by employing a two stage hybrid localisation framework. First, a convex Semidefinite Programming (SDP) relaxation is used to find a globally consistent initial estimate, which is then refined through a non linear Gauss Newton Maximum Likelihood Estimation (MLE). Field experiments in varying environments, including dense forests, show that the system’s performance is highly dependent on receiver geometry and data quality. A key result is the implementation of a node removal strategy based on residual and sensitivity metrics, which improved localisation accuracy by up to 90% in certain scenarios. Additionally, the study explores the use of bearing intersections between receiver clusters to maintain tracking capabilities even in severely degraded conditions where absolute signal accuracy is low. These findings demonstrate that RSS localisation is a viable and scalable solution for tactical monitoring where flexibility and low cost are prioritised over specialised infrastructure.