Unlocking AI in Space

Thousands of satellites are being launched each year on the back of advances in technology and a diversifying space industry. Among the more conservative estimates, Euroconsult expects 2,800 launches annually through to 2032, to meet the growing demand for global connectivity, Earth observation, and space exploration.

And as in numerous other industries, operators are using artificial intelligence (AI) to address the dynamics of busier orbits. Machine learning algorithms, neural networks and computer vision are among a host of AI technologies that are being used to optimise and automate operations, analyse data, and expand industry use cases – not least to the growing demands for satellite connectivity.

The International Telecommunication Union (ITU) offers some examples of how AI can tackle challenges across many aspects of the satellite ecosystem in a recent publication: In the maritime and aviation sectors, for example, satellite information and capabilities can be combined with IOT sensor data to enable streamlined operations for shipping and airline activities. AI will facilitate next-generation satellite systems to be autonomous, with a dynamic resource management that will be capable of adjusting coverage, capacity and spectrum as needed.

But perhaps the most famous of AI applications is NASA’s Mars rover. Including spacecraft operations and on-board data processing, the space agency is estimated to have spent $233m on AI applications between 2020 and 2022, according to analysis by project software platform Deltek.

Other players, including Boeing, Lockheed Martin, Space X, Thales Alenia, Airbus and Amazon’s Kuiper Systems, likewise deploy a clutch of different AI technologies.

The UAE’s newly formed SPACE42 is going one step further, building its entire operations around AI. Created in December by the merger of Abu Dhabi companies Al Yah Satellite Communications Company (Yahsat) and Bayanat AI, the new entity aims to capture regional and international opportunities in satellite communications, geospatial and mobility solutions, earth observation, internet of things (IoT) applications, and business intelligence.

Yahsat contributes advanced satellite communication capabilities while Bayanat brings geospatial analytics expertise to SPACE42, Managing Director designate Karim Sabbagh tells SatellitePro.

“By combining Bayanat and Yahsat, we will have a unique positioning regarding space assets—owning from the ground, stratosphere, LEO, and GEO orbital layers. This will enable SPACE42 to tackle issues with unmatched precision and efficiency, streamlining the end-to-end process and allowing for unparalleled insights,” he says.

Its creation reflects the UAE’s aim to transform the economy into an innovative, AI-driven ecosystem in line with its National Space Strategy 2030, he added.

Bayanat and Yahsat have a combined revenue of Dh2.8bn ($7.6m), with a net income of Dh637m ($173m) for 2023, according to a circular released in March. SPACE42 is expected to begin operations this year.

Speaking of SPACE42’s addressable market, he says: “First, there is a growing need for use cases that combine geospatial intelligence and satellite communications. Second, there is a growing opportunity to intertwine the operations of space and ground systems across geospatial intelligence and satellite communications. Finally, AI will be crucial in powering the differentiated use cases we want to serve, and optimising the operation of space and ground systems.”

AI could widen the playing field in the space and satellite sector, shorten development cycles and reach more people with a greater depth of applications, explains Sabbagh. Potential outcomes include using AI analytics to multiply technology productivity, such as by improving data transmission and service quality for revenue growth, to extend the life of space infrastructure through proactive maintenance. AI in system design and automation could reduce mission costs and make space-based services more accessible and sustainable, potentially improving throughput by 10x. Third, shorter technology cycles become possible, such as by way of AI-powered digital twins in manufacturing small satellites or midstream multi-asset system management.

Bridging connectivity gaps

Internet connectivity is particularly important for job creation and business growth, two areas where the MENA region can benefit, the World Bank pointed out in a regional analysis. Satellite internet could bridge the infrastructure and funding gaps facing the region when it comes to better internet access.

Satellite-based internet connectivity could extend broadband coverage universally, narrowing the digital gap and uplinking connecting remote industrial facilities, while enabling a wide range of new capabilities and business models.

It’s this market that the companies such as SpaceX’s Starlink, Eutelsat Oneweb and Amazon’s Project Kuiper are looking to address.

In December, Amazon Web Services CEO Adam Selipsky announced plans to offer private networking capabilities alongside Amazon Web Services (AWS). Using its Low-Earth Orbit (LEO) network, the connectivity service targets business, enterprise, operator and public sector customers to bypass the public internet and move data privately into any cloud region. Amazon has announced a collaboration with Vodafone and Vodacom to deliver 4G and 5G services to customers in Europe and Africa. A similar partnership with Japanese companies announced in December highlighted how Project Kuiper would enable customers to connect to Amazon’s cloud-based infrastructure to use AI and ML applications more widely, while expanding off-the-grid tracking and monitoring Internet of Things (IoT) devices.

But although demand is growing for anytime-anywhere connectivity, delivering internet from space carries similar cost and complexity constraints as terrestrial internet networks. AI can reshape satellite connectivity by enhancing network performance, efficiency, and accessibility, leading to more efficient operations and speedier returns on investment across the board.

Likewise, AI algorithms can help optimise networks, enabling satellite operators to bridge connectivity gaps in areas that are underserved by traditional infrastructure (including remote industrial installations). Similarly, dynamic bandwidth allocation enables operators to open up – periods to new customers or applications. With lower latency and better coverage, satellite clients get the connectivity they need on demand – especially in remote or underserved areas. Enterprise customers, in turn, can harness the real-time data-crunching abilities of machine learning (ML) applications such as large language models (LLMs) together with cloud services to maximise bandwidth consumption, expand their customer base and speed time to market for new services.

Likewise, AI can help operators predict and reduce signal interference, leading to more reliable communication links.

“Optimising space connectivity through AI is expected to be critical in furthering our efforts to help unserved and underserved communities,” SPACE42’s Sabbagh says.

Besides furthering regional satellite connectivity, the new company wants to reach communities beyond the Middle East, he notes. Yahsat’s satellite fleet already delivers communication services such as broadband and video broadcasting to over 80% of the world’s population in Europe, Asia, Africa, and Australia. These services will continue, including with the $500m Thuraya 4-NGS, mobile telecommunications system to expand Thuraya’s L-band business for customers including defense, government and enterprise.

Improving business intelligence

Satellites generate vast amounts of data on a regular basis, principally from Earth observation, and for weather forecasting, but also for espionage and increasingly, environmental monitoring.

The US National Oceanic and Atmospheric Administration’s network of satellites and Earth-based observation system together collect some 20 terabytes of environmental data every day.

Machine learning can quickly mine these large datasets for intelligent and actionable insights. As we have seen with ChatGPT, for example, LLMs can exponentially improve data processing and analysis speeds.

With machine learning, satellites can automatically detect, classify, and interpret objects and events, providing valuable insights for disaster response, agriculture, urban planning and resource management. Additionally, AI enables predictive analytics, spotting trends and anomalies in satellite data to help decision-makers make proactive choices and manage risks effectively.

Worldwide, the market for AI in remote sensing and earth observation could be worth $35.90bn by 2030, up from $1.75bn in 2022, industry estimates show. That’s a compound annual growth rate of 27.5% through to the end of the decade.

In February, the US-Saudi joint venture SpaceGuardian became the first company dedicated to satellite imagery and AI analysis to launch in Saudi Arabia, bringing critical intelligence-gathering capabilities to the region. In partnership with New York-based SpaceKnow, SpaceGuardian will use ML algorithms to provide business insights in the energy, defence, infrastructure and environmental sectors, among others.

“We aim to revolutionise the AI-powered geospatial analytics market and to bring our technology to multiple sectors, starting with infrastructure monitoring, carbon sequestration, and security applications,” says Jerry Javornicky, CEO and co-Founder of SpaceKnow.

The collaboration will support Saudi Arabia’s economic diversification efforts and support a growing space sector, adds Majid Alghaslan, Chairman and CEO of Front End Limited Company and Founder & Partner of SpaceGuardian.

In the meantime, Florida-based Sidus Space goes one step further, using both hardware and software AI across its products. The vertically integrated Space and Data-as-a-Service company has business activities spanning hardware manufacturing, multidisciplinary engineering services, satellite design, production, launch planning, mission operations, and in-orbit support.

Sidus’ FeatherBox AI computing module onboard its LizzieSat satellites processes data directly from onboard sensors and only transmits the relevant information to customers. This cuts down downlink transfer costs and speeds up responses to observed events.

“With our FeatherEdge hardware/software AI/ML data processing device that sits onboard each of our LizzieSat spacecraft platforms, we can process the data that we collect in near-real-time to reduce data footprint and transmit intelligence with lower latency,” Carol Craig, founder and CEO at Sidus Space tells Satellite Pro.

Currently, all data is downlinked to the ground before processing to determine if there is useful data, which can take hours or even days. “We aim to deliver insights to end users within 60 seconds, which will enable a broad array of time-sensitive applications,” she elaborates.

“Performance-wise, we enable our customers to receive information on their area or event of interest in time-critical use cases. These use cases include illegal fishing and other marine domain awareness activities, which are of particular interest in the MENASA region,” she remarks.

The analytics solution improves Sidus’ business prospects, enabling it to offer tailored intelligence solutions for industries such as defense, agriculture, maritime, and oil and gas.

“We’re focused on the platform that enables the implementation of complex AI/ML data processing applications in the resource constrained satellite environment. We can then host data processing applications and update them on orbit to provide a flexible Platform-as-a-Service solution that enables our customers to rapidly adapt to changing mission needs,” Sidus says.

“In the future, our onboard AI could be used for smart routing and cross-linking between satellites to enable satellite connectivity and efficiency,” she adds.

Automating satellite operations

But perhaps AI’s most transformative impact in the satellite sector is in making satellites smarter by advancing autonomous operations, reducing reliance on ground control and enabling greater autonomy in mission planning, execution, and maintenance. Just like Sidus’ data analysis systems, onboard AI diagnostics and predictive maintenance algorithms can track satellite health, forecast and mitigate technology failures to minimise downtime and extend satellite lifespan.

AI-powered automation can streamline satellite operations, reducing the need for human intervention, cutting operational costs and response times, and increasing mission efficiency and reliability. With more dependable satellite services, operators can optimise their investments in space technology.

Efficient automation and ML/ AI processes have long been helping achieve a level of precision crucial for ensuring the safety and integrity of satellites, Helen Weedon, Managing Director, Satcoms Innovation Group, tells SatellitePro. “ML and AI processes, both on the ground and in the satellites themselves, are already ensuring simultaneous control of over thousands of LEO satellites in space, without a single major incident for several years. Thanks to these technologies, we are now able to spot conjunctions with a distance as small as 500 meters, at more than 10 km/s velocity difference at 550 km altitude.”

Beyond safety considerations, applying AI and ML has the potential to sharply increase efficiency in a number of satcom applications. “This is especially possible wherever we have high quality and accurate data available, e.g. coming from sensors via telemetry from the satellites themselves. By adding this to high quality external data, it is possible to complete important tasks such as satellite orbit calculus or payload incident detection,” Weedon adds.

Last year, Lockheed Martin unveiled a new advanced satellite operations centre near Denver, Colorado, to test how AI, automation and cloud capabilities to manage the increasing number of satellites constellations in LEO orbits – with fewer staff in near ‘lights out’ conditions. The facility could enable simultaneous control of multiple missions and remote access for operators worldwide thanks to using consistent software across missions through a web-enabled secure cloud framework. With existing functionalities incorporated into the facility, a single operator could potentially oversee individual satellites and entire heterogeneous constellations of different kinds of satellites from nearly anywhere over the internet, says Maria Demaree, Vice President and General Manager of Lockheed Martin Space’s National Security Space business.

Navigating AI’s challenges

Again, as in every other industry, satellite operators face significant challenges in adapting to AI and embracing its gifts requires careful consideration.

Weedon warns against viewing AI as a cure-all solution. Although satellite operators have huge volumes of machine-generated data in their database ready to be used to increase efficiency, doing so requires strategic leadership and organisational adaptability – particularly when using the technology to improve existing business functions. “Accepting and implementing change, including a well-defined risk management, is extremely important through all levels of employees,” she says.

Weedon flags the need for planning in the automation process. Early-stage development requires careful consideration of scope, efficiency, and process gains, she adds, pointing out that automation attempts without a clear understanding of data usability can lead to project failure and wasted investment.

“Any later change will be immensely complicated to re-implement, requires often a start over from square one.”

She says implementing AI systems can be both complex and iterative in nature. In the rush to secure a quick win, companies can struggle with the technologies’ mathematical intricacies, with the project often ending in failure.

“ML/AI requires disruptive thinking and is not easy to implement, neither in workload, nor in understanding, nor in cost,” she says. “All that we see today at ChatGPT, and other AI applications need years of development without having a single cent of return of investment.”