GPU-accelerated Direct Geolocation of GNSS Interference

TL;DR

This paper introduces a GPU-accelerated method for direct geolocation of GNSS interference, significantly reducing computational time and enabling real-time detection and localization from LEO satellites.

Contribution

It proposes a parallelized GPU implementation of direct geolocation, overcoming computational challenges for LEO-based GNSS interference detection.

Findings

GPU acceleration greatly reduces processing time

Parallel processing enables real-time interference localization

Method outperforms traditional CPU-based approaches

Abstract

In recent years, there has been a sharp increase in Global Navigation Satellite Systems (GNSS) interference, which has proven to be problematic in GNSS-dependent civilian applications. Many currently deployed GNSS receivers lack the proper countermeasures to defend themselves against interference, prompting the need for alternative defenses. Satellites in Low Earth Orbit (LEO) provide an opportunity for GNSS interference detection, classification, and localization. The direct geolocation approach has been shown to be well-suited for low SNR regimes and in cases limited to short captures -- exactly what is expected for receivers in LEO. Direct geolocation is a single-step search over a geographical grid that enables estimation of the transmitter location directly from correlating raw observed signals. However, a key limitation to this approach is the computational requirements. This computational burden is compounded for LEO-based receivers as the geographic search space is extensive. This paper alleviates the computational burden of direct geolocation by exploiting the independence of position-domain correlation across candidate points and time steps: nearly all computation can be accomplished in parallel on a graphics processing unit (GPU). This paper presents and evaluates the performance of GPU-accelerated direct geolocation compared to traditional CPU processing.