Advancements in space technology continue to push the boundaries of possibility for orbit, with spacecraft autonomy seen as the key to next-generation space systems. These new systems promise to enable new applications and increase the efficiency, responsiveness and continuity of space-based Earth and science observations, communication and connectivity, meeting our Defence and National Security needs. The aim is to fly smarter spacecraft, with a reduced burden to on-ground operators and users and built around a responsible approach in congested orbital regimes

The potential of spacecraft autonomy for Australia

For Australia, the possibilities are particularly exciting. While the nation’s space sector is still relatively young, it is already producing game-changing technologies, including world-firsts across many sectors, from defence to communications. Distributed architectures and constellations of small spacecraft are achievable prospects that will enable us to take advantage of these new space systems.

One key factor in realising this vision is spacecraft autonomy and onboard artificial intelligence (AI). Recognised as a key capability, this is an opportunity for Australia to build on its technology prowess, developing these innovative technologies and contributing to the international space ecosystem. This is a huge opportunity for Australia to develop Al-based spacecraft autonomy capabilities for the next-generation space systems to enable national missions and achieve home-grown benefits of remote sensing, communications, and other vital services from space.

Spacecraft autonomy can cover many spacecraft and payload aspects, from planning and executing missions on board to controlling the spacecraft in real-time, even when out of contact with ground stations. Key applications of spacecraft autonomy include:

Continuous mission planning & real-time execution

One of the primary benefits of spacecraft autonomy is the ability to perform continuous mission planning and execution onboard without relying on constant communication with ground stations. This allows the spacecraft to adapt to changing conditions and respond to unexpected events in real-time without needing to wait for commands from Earth. This can include an ability to navigate autonomously and adjust its trajectory based on scientific observations, avoid hazards or conjunctions or meet mission requirements.

Maximising mission objectives

Onboard decision-making enables the spacecraft or its payload to maximise mission objectives by utilising the available onboard resources or capabilities of nearby constellation spacecraft for edge processing. For example, if a spacecraft has limited power or fuel, it can prioritise its activities to ensure the most critical obiectives are achieved before exhausting its resources. Additionally, spacecraft can work together in a coordinated manner to achieve shared goals, such as performing scientific observations or communicating with Earth.

System robustness

Spacecraft autonomy also enhances the robustness of the spacecraft systems, which refers to its ability to continue functioning in the presence of onboard failures and external uncertainties. By implementing fault-tolerance and self-healing mechanisms, spacecraft can detect and recover from failures autonomously, reducing the risk of mission failure or loss.

SmartSat is supporting the move towards resilience through a “smaller, smarter, many” spacecraft approach, focusing on technologies to enable these future distributed space systems and small satellite constellations. This hybrid space architecture is a shared vision with Defence allies, while also applying to future commercial and science-based constellation missions. Autonomy is a critical enabler for resilience from space, and for ensuring future Defence services remain resilient during threats, faults, and unexpected circumstances. SmartSat researchers have defined five target outcomes as a focus for delivering:

• Smart sensing: Obtaining “live” information and reduced-latency actionable intelligence, including direct user tasking through onboard processing for realtime inferences
• Smart resiliency: Protecting and operating assets in a contested and congested environment, with local awareness, fault and threat protection and responses for resilience
• Smart spacecraft: Squeezing more capability out of small spacecraft and their payloads through management and optimisation of onboard and distributed resources
• Smart constellations: Federated constellations and operations with minimal human interaction, providing services to the user (as opposed to operating a spacecraft for a service)
• Smart assurance: Building autonomy around trust and ethical choices, so actions are predictable, and confidence is quantified to the user/operator

Achieving true autonomy in space flight

The adoption of truly autonomous spacecraft has been limited due to the inherent risk-averse nature of operating a complex and expensive spacecraft in the remote and unforgiving environment of space. As a result, large operations teams and onboard state machines are still the norm. However, the space industry is adopting edge processing and payload intelligence to overcome these challenges to fully realise the potential of spacecraft autonomy.

One of the areas that needs to be addressed for true autonomy is the size and weight constraints of small satellites. These also offer an ideal platform for training, running and maintaining large deep learning (DL) models on low-GPUs or single-board computers. Small satellites also offer the potential to optimise the design of DL models to minimise power usage and memory footprints, without impacting overall performance.

Another area to explore is how deployed Al/Machine Learning (ML) algorithms should operate in edge scenarios without additional support from the ground stations. For example, considering how to recalibrate deployed DL models in orbit by detecting the data drift issue and re-developing models by adjusting the architecture and parameters. It is also vital to examine the inherent limitations or minimum DL building blocks needed to extract actionable intelligence onboard for specific end-user applications.

Achieving autonomy in space demands real solutions to all these fundamental and practical issues related to onboard processing and trusted Al.

Recent progress in the field

Spacecraft autonomy has been actively researched for some decades. The implementation depends on specific mission requirements and constraints, varying between a very low level of autonomy with a high level of ground control, to a high level of autonomy where most functions are performed onboard.

Most current Al-based solutions are typically deployed with the main processing performed on the ground, while onboard processing often limited to data collection, filtering, and aggregation. Recently, there has been an increased focus from academia and industry on generating more actionable intelligence onboard. Several companies are now providing commercial software for Al-based onboard processing, including the AIKO’s MiRAGE solution for autonomy (onboard data processing for perception) and downlink optimisation through onboard image processing.

Another area of progress is fault detection, isolation, and recovery (FDIR). Functionality has been gradually transferred from the ground to the spacecraft to ensure safe and high-performance operation with less intervention by human operators. This is particularly important for deep space missions or missions involving hundreds of small spacecrafts. However, FDIR uses pre-defined recovery strategies and cannot anticipate potential threats, unexpected errors or malicious attacks.

To this end, there is a need for Australia’s space sector to develop novel concepts, methods, and technologies to provide new Al-based spacecraft autonomy capabilities for next-generation space systems. Selected areas of interest to SmartSat include onboard processing and actionable intelligence, small spacecraft and constellation resilience, dynamic optimisation of constellation resources and real-time tasking and resource allocation. Through its broad program of satellite research, SmartSat is well placed to take on the task of progressing Australia’s spacecraft autonomy capabilities.

Australia is building skills, hardware, and capability in its burgeoning space industry and this must include autonomous algorithms to remain at the innovative front of space system development and implementation. As autonomous technologies advance and are embraced, their place in space will expand and present new opportunities for emerging Australian space technologies. The more we push the boundaries of what is possible, the more exciting the future of space exploration begins to look. With next-generation space systems on the horizon, the possibilities for science, discovery and innovation are endless.

By Dr Carl Seubert, Chief Research Officer, SmartSat CRC and Ryszard Kowalczyk, Professorial Chair of Artificial Intelligence, SmartSat CRC/University of South Australia

This article was first published by Australia in Space magazine