Trends and Applications of AI in Space

For several years, the satellite and commercial space sector has sought ways to automate equipment construction, foster innovation, and boost profitability. Artificial intelligence (AI) can support these efforts. More specifically, the technology can change the way global satellite operators and space agencies process data and transform how the sector operates across four key areas: manufacturing, imaging, telemetry, and spectrum usage.

Manufacturing

AI has the potential to significantly improve the satellite manufacturing process, particularly when meticulous engineering is required to assemble multiple pieces. Newly developed AI technologies can perform tedious, time-consuming yet necessary tasks, such as cleaning satellite parts. Measurements can be gathered automatically and updates on the health status of core components can be easily shared with engineers. Such an application of AI will not only generate profit, it will reduce production time, allowing satellite operators to launch sooner than before.

Enhancing Earth Exploration and Imaging

Agencies and governments can leverage AI technologies to gather precise earth exploration data. Robotics can be used to identify areas of surveillance by learning to process and act upon signals it receives, disregarding substantial amounts of unnecessary data. According to the European Space Agency (ESA) satellites can provide over 150 Terabytes (tb) of data per day. AI would reduce costs, extend mission and battery life, and produce higher quality environmental image data. Earth imaging data is already being used to provide actionable insights for governments and businesses; calculating macroeconomic activity to more accurately measure migrant flows and the impact of climate change.

Operations, Telemetry, and Control

AI is being used by some to monitor telemetry and provide feedback to control satellites. For example, SpaceX has implemented AI operations to avoid satellite collisions. But the technology could be used for other tasks, such as executing debris avoidance maneuvers automatically. While it may provide part of the solution to the space debris issue, some experts have raised concerns over the necessity of ephemeris data sharing between operators. The Satellite Innovations Group, Airbus, and the Space Data Association have been researching applications of these techniques.

However, the widespread use of AI increases the risk of unauthorized system hacking, including the manipulation of software leading to signal blocking, satellite takeover and destruction. But while it creates new threat vectors, AI can unlock protective cybersecurity applications that allow operators to stay one step ahead of malicious actors.

Dynamic Spectrum Detection and Avoidance

Another application of AI in the space industry is dynamic spectrum usage. A satellite can learn to transmit using the appropriate frequencies and level of power output to avoid interference. Deep learning can be used for space-to-Earth transmissions and has wide-ranging implications for simplifying coordination. Currently, Wi-Fi uses a form of Dynamic Spectrum Access (DSA) which could be used in a new generation of satellites. According an IEEE report, the technology used for RLAN can be enhanced to lower the chance of interference and increase spectral efficiency. However, any technological application of Dynamic Frequency Selection (DFS) may take several years to rollout due to the average lifespan of satellites and spectrum changes will require international regulatory change and compromise within the ITU.

These dynamic technologies may have particular value in the telemetry and control of Geostationary Orbit (GEO) and Non-Geostationary Orbit (NGSO) frequency and physical coordination. Such procedures to eliminate interference are not always effective; GSO operators including Viasat have reported extensive interference from NGSO systems and numerous algorithms are required to manage transmissions and coordinate spectrum before any transmissions occur. However, adopting deep learning technology and automatic detection of transmitted frequencies from networks in proximity will reduce the interference burden for satellite networks. Technology can learn to detect and avoid co-channel interference at different stages of the satellite orbit. Some form of deep learning is envisaged as the growing number of NGSO constellation becomes increasingly challenging to manage.

Benefits and Challenges

AI can provide numerous benefits to society in the context of space. Potential developments in manufacturing, spacecraft control, and coordination would ease the challenging constraints on operators, governments and agencies as well boost productivity and growth in the downstream satellite industry. AI can drive down consumer costs, foster innovation, provide solutions to the issue of space debris, and guarantee certainty for operators who find network coordination increasingly complex.

The implementation of AI for space is not without its challenges. Domestic legislation and international regulatory frameworks are not particularly favorable and supportive of AI. For instance, ensuring dynamic coordination requires confidential information to be shared between operators so machines can develop avoidance mechanism. Current international data retention laws are unlikely to allow such an exchange of data.

Increasing transfers and storage of data in space via satellite may also create regulatory hurdles. With countries adopting territory-based system for data governance, questions arise pertaining to what rules should be adopted to govern transfer, retention, and access processes in space where no country has a territorial claim. Issues of territoriality and governance may also become more significant as governments move towards regulating AI on issues of accountability and transparency. VS