This project aims to develop a novel distributed radio frequency processing capability for deployment across space-based platforms. The research will focus on elaborating a machine learning (ML) model that can effectively detect and classify conventional and low-probability-of-intercept (LPI) radar signals. The model will dynamically adapt its structure to the collected signals, producing calibrated confidence estimates and learning from misclassified in-distribution samples. To enhance autonomy and resilience to mission contingencies, the neural network will perform asynchronous, distributed training featuring
interruptible and resumable cycles.

Satellite imagery provides immense potential for Earth observation applications. By training AI to analyse these images, we can monitor huge areas with minimal effort. However, there are many technical hurdles that need to be overcome to successfully use these images for application areas such as agriculture and ecology. Satellite imagery has much lower spatial resolution than ground-based images, and the higher resolutions satellite images (30cm) must be taken selectively as it is infeasible to capture the whole planet at that resolution. This leads to sporadic images being available for any site with only one image taken every few months, or several on the same day. To train an AI model on these images, we need to construct training sets of ground-based observations associated with positions in satellite images. The ground truth observations are also sporadic, and these combine to make training an AI algorithm on satellite images challenging due a lack of observations that are temporally aligned with images.

This project aims to develop sample efficient AI algorithms which require less ground truth observations to train accurate models. Both semi-supervised and unsupervised training methods will be adopted to train highly effective feature extractors using a minimum of labelled data. Highly successful semi-supervised learning algorithms, and unsupervised/self-supervised methods (which have achieved very high accuracy with just a few labelled examples or no labels at all (respectively) in generic image classification) will be customized and these methods extended for satellite images.

Distant clock synchronisation has many applications including telecommunications, Global Satellite Navigation Systems, emitter localisation, and phased array astronomy. Clock synchronisation requires clocks to tick at the same rate and to read the same time. Two procedures employed to ensure clocks read the same time are the Einstein method and the Eddington method.

The Einstein method involves sending a signal back and forth between two clocks and using the speed of the signal to synchronise the clocks.

The Eddington method involves synchronising the clocks next to each other prior to sending the second clock to its desired location.

Special relativity and general relativity must be accounted for in the Einstein and Eddington methods to ensure that the clocks run at the same rate. Furthermore, the Eddington method relies on the physical movement of a clock and is therefore not practical for space applications. This project will investigate the employment of entangled photon pairs in overcoming the limitations of the Einstein and Eddington methods.

Precision timing is a key capability underpinning the operational efficiency of society’s most critical infrastructure and is making new applications possible. Global Navigation Satellite System (GNSS) satellites carrying atomic clocks have contributed to delivering a global timing capability with high levels of accuracy and stability. The increasing demand for GNSS independent timing solutions, as well as the potential for optical clocks offering higher levels of timing accuracy and stability demands a study to address the potential applications and technologies required for success in these areas.

This project will investigate what new opportunities and resilience a compact, high accuracy clock for use on small satellites would enable for a broad range of precision timing applications.

Extensive research in Geostationary orbit (GEO) failure modes optimise resilience of single satellite systems. Resilience measures of Low Earth Orbit (LEO) constellations failure modes and their optimisations are less well understood.

This project concept will deliver a technology horizon scan for resilient, efficient and effective management of constellations of LEO satellites as well as a suggested roadmap for areas of focus for the SmartSat CRC.

To truly exploit the advantages of Distributed Satellite System (DSS) mission architectures, an evolution is required from the inflexible pre-planned approaches of traditional space operations to systems that are suited to reactive and resilient mission architectures. At its core, this requires the design and development of novel intelligent Mission Planning Systems (iMPS) that facilitate autonomous Goal-based Operations (GBO). From a technical standpoint, iMPS must facilitate the autonomous cooperation of DSS to optimally global systems goals within an uncertain, dynamic mission environment. From the human perspective, GBO marks a paradigm shift from a command sequence role to one of a supervisory nature, where system autonomy must be monitored and managed in near-real time.

This research explores the concept of supervisory control through the design and development of an human centric iMPS for autonomous GBO. This system will enable an operator to express their intentions in the form of system goals, project and visualize the effect of these intentions and provide intelligent mechanisms to curb autonomous system behaviour. System design and development will follow a Model-Based Systems Engineering (MBSE) approach and verified through case studies that include bushfire detection and maritime surveillance while considering key dynamic mission aspects such as the availability and throughput of communication (e.g., inter-satellite laser links) systems.