Parallel Processing for Novel Navigation

This project focuses on exploring the power of GPUs and multi-threading CPUs to enhance particle filters for use in inertial navigation systems. GPS jamming technology is becoming more sophisticated and therefore there is a growing need for novel navigation solutions. Inertial navigation systems are currently too sensitive to acceleration measurement errors, which can allow drift in the position measurement, rendering the position inaccurate over long durations of time. The use of a particle filter offers a promising solution to mitigate these challenges and enhance the robustness of inertial navigation systems. This project will be aimed at making best use of modern processing hardware (GPUs, multi-threading CPUs, etc) to improve the accuracy and robustness of a multi-sensor Positioning-Navigation and Timing (PNT) system. It will also cover distributed processing methods, sensor fusion, navigation and timing systems, and will provide opportunities to gain experience with real-time processing with sensor hardware.

The potential for using parallel implementations of sensor processing for PNT applications is huge. Some aspects, such as the sequential nature of inertial navigation processing, are not amenable to parallelisation, but others are more compatible. For example, the use of 'particle filters' (also known as sequential Monte Carlo methods) for sensor fusion have great potential, and the University of Liverpool is a centre of excellence in this area. Recent work at Liverpool has demonstrated streaming (online) particle filters that can be parallelised to obtain the full benefit of multiple processors. The key objective of the PhD project will be to make the full benefits of parallel processing available to a multisensory PNT system.