Radio Source Signal Localization Methods and Hardware Simulation

Abstract

This thesis identifies and derives several methods of transmitter localization as well as collects several pre-existing algorithms and adapts them to the use of passive GPS interference localization. Due to legal transmission limitations, the use of Anti-Jam algorithms in a custom front-end architecture providing protection to a GNSS and INS navigation sensor is simulated via hardware in the loop (HWIL) to mimic hostile navigation environments. The HWIL sys- tem consists of multiple calibrated software-defined radios (SDRS), propogation delay systems, GNSS/INS systems, timing, network, and processing devices. Each controlled by custom or open-source software to create a system capable of simulating transmitted signals as would be seen in a wavefront scenario. This system demonstrates moderately realistic signal source lo- calization. This conceptual platform, designed to observe and localize a signal, requires several sensors for observing its own location in the challenged RF environment. The suite of sensors on this platform would include a GNSS and INS system to provide both position and attitude that utilizes a fixed array of antennas. The array, known as a controlled reception pattern ar- ray(CRPA), is designed for the navigation band of choice as well as the signal of interest. The CRPA, utilizing mulitple signal classification (MUSIC), provides a direction of arrival (DOA) to the transmitting source. The combination of the platform position, attitude and DOA collec- tively provide the required information to generate estimates of the transmitter location. The estimation methods referenced and derived include batched and continuous estimation tech- niques utilizing the combination of the available observables on the theorized platform. The use of closed form solutions with the noisy measurements are used only as initialization points for the batch and continuous estimators. The estimation is done in two and three dimensions for various esimators. These methods are selectivly compared based on the applied estimation technique. The comparisons are divided into two groups, stochastic and instantaneous. The instantaneous group is used as a method of initialization of the stochastic filters but should also ii be considered as an estimate method in multiple observer cases. The centroid algorithm pro- vides moderately better results than other methods reviewed in the simulation. The stochastic solutions are compared in the second group. The maximum liklihood estimator provides the best solution consistently. While the particle filter method and extended kalman filter provide similar results, they fall victim to different failure modes induced by the observer path.