Artificial intelligence has caught the interest of businesses, government, defense, and the general public alike, each group imagining the possible benefits to everyday life and the potential for the future. Satcoms is no exception, with entities throughout the value chain watching closely.
Helen Weedon, Managing Director, Satcoms Innovation Group
For some time, we have been discussing the potential of Artificial Intelligence (AI) to revolutionise the Satcoms industry, from detecting problems to maximising efficiencies. While it has not yet reached anywhere near its full potential, it does seem as if some progress is being made, both in terms of the technology itself and how that is applied to Satcoms. What is the full potential of AI for Satcoms and how do we make sure we harness it effectively?
AI and machine learning
When we talk about AI, we are more often referring to machine learning, essentially teaching a machine to do something by example. Every time your Internet browser asks you to identify which squares have a certain image in them, you are adding to the machine’s knowledge. AI can actually apply to a whole range of different techniques. One such example is knowledge engineering. This is about extracting the knowledge of those experts in the field into a set of generic rules and algorithms that can be applied for problem solving, something that is challenging but could be very interesting in the Satcoms industry.
The use of AI across a multitude of different sectors has been steadily increasing. It is already being used for a range of applications, from serving customers with personalised ads and content recommendations, to analysing chronic conditions with other medical data for early diagnosis. With each new application, the underlying technology is getting better and better. And all the reports seem to indicate that it is set to rise further over the coming years. According to Statista, the global AI software market will reach US$22.6 billion by 2025, with approximately 54 percent growth year-on-year. Gartner forecasts that 50 percent of enterprises will have devised AI orchestration platforms to operationalize AI by 2025. This is up from fewer than 10 percent in 2025. The same report also states that AI will be the top category driving infrastructure decisions. Gartner also believes that by 2024 the degree of manual effort required for the contract review process will be halved for those companies using advanced contract analytics solutions.
Why AI makes sense for Satcoms
At the same time, the Satcoms industry has been growing rapidly over recent years. We have never had more satellite launches than we have had in the last two years with all the LEO launches. New launches and increased ground equipment are all adding complexity to an already challenging environment. As the skies become more crowded, there is an ever-increasing risk of things going wrong. Whether that is satellite interference or large-scale collisions, the consequences will be dire.
Anything we as an industry can do to reduce those risks has got to be worthwhile and AI has the potential to do just that. At the same time, one thing we are not short of in this industry is data, the cornerstone of any AI solution. Coupled with the presence of lots of manual, repeatable processes, it is the perfect recipe for AI to have maximum impact.
However, understanding present risks and even more judging upon and using AI in the right way to apply effective countermeasures is a demanding task. It is not something that can simply be applied quickly and forgotten about. We need to think about probabilities of events which might happen, are detected and might be eventually mitigated. AI cannot take out all remaining risk percentages and will never be able to do so, but it can help the industry to understand that challenge a lot better. The key to have these values calculated well enough to use them lies in the definition of questions, answers, and recurrent acting advice by senior expert engineers, as well as the acceptance of any commercial and environmental risk by the company decision makers. Implementing AI is not a task that engineering or management can decide upon alone. Both groups need to be in full understanding if we are to have AI implemented successfully in our industry.
Erwin Greilinger, Product Line and Sales Manager for Satellite Monitoring Solutions, at Atos, commented: “AI has the potential to be a game changer in the satellite industry, where it could be used to improve processes and increase efficiency. AI could establish a digital assistant, capable of processing information about everyday satellite procedures, resolving and predicting issues before they become errors. On top of this, it could be taught a whole spectrum of things relating to the day-to-day tasks of satellite operations engineers.”
AI development so far
As mentioned above, we are beginning to see innovation for AI applications within the Satcoms industry. One area in particular where there has been lots of innovation is that of onboard observation and data processing. This includes from companies such as LatConnect 60, Raytheon Intelligence & Space, CS3, and Northrop Grumman. Building AI tools into the satellites themselves so that they can process data collected in space and alert operators to problems early seems like a leap in the right direction. In LEO, AI and ML are much more prevalent, being used already for orbit prediction by current operators, for example.
There have been several other innovations, from using AI in manufacturing, to predicting the location of satellites and debris and monitoring the health of satellites in orbit. There have also been a number of AIspecific research projects undertaken, such as from the Centre Tecnològic de Telecomunicacions de Catalunya, which has been comparing RF signals for interference detection using Signal Correlation, ML and AI. Having demonstrated automated GSM interference retransmission detection based on I/Q samples, work has been done in projects to identify also other types of interference correlation such as ASI and XPOL, making it possible to have interference detection and analysis done in the background without first human intervention.
SIG member, Integrasys, sees itself as pioneers in AI within the Satcoms industry. Its product portfolio drives end-to-end automation, taking previously manual tasks and making them automated. Alvaro Sanchez, CEO, Integrasys, commented: “We have an extended experience in network automation solutions. Our product portfolio covers design and deployment to monitor and maintenance, with no human touch, intuitive interfaces, and high-quality outputs that help to transform space data into tangible operations in ground by an intelligent system.” He added that its latest patents focus on AI advanced methodologies for end-to-end automation in the new LEO and MEO world.
Atos is another member that has been extremely active in the development of AI tools and technology. Throughout 2020, Atos carried out extensive research and development looking at how AI could be applied to carrier monitoring and geolocation systems to address a number of issues and improve the efficiency of these systems. Greilinger said: “To combat the rising interference challenge before it becomes completely out of control, new solutions are being developed. While working groups and associations like the Satcoms Innovation Group (SIG) and the Global VSAT Forum (GVF) educate a growing number of members on both proactive and reactive responses, the best cure for satellite interference has been and will continue to be vigilant monitoring by advanced sophisticated interference detection systems, with algorithms that determine the identity of interfering carriers.”
The research carried out by Atos has proven that AI can help to address interference. It was able to enable automatic identification of interference situations via spectrum analysis, predict satellite link quality degradation caused by bad weather scenarios, and enable interference detection in complex satellite environments.
The future AI use cases
As technology continues to improve, the potential use cases in the satellite industry are extensive and far-reaching. I can see a future where AI will feature right through the satellite chain in all orbital regimes, from autonomous operations for satellites, to automated deployment of ground infrastructure, to interference detection, classification, and prevention. It could also have a huge role to play in space situational awareness and collision avoidance.
It is certain that getting there requires a new technology to be taken up by satellite operators – mathematics. After RF as a baseline and IT on top, this will be the third main domain and a revolution in technology for conservative GSO satellite operators. This will be much quicker for new space operators coming to the industry with a fresh perspective.
Satellite engineers will also be critical to the success of AI. Partly that is about tapping into the extensive knowledge of today’s engineers. Perhaps more importantly, it is the engineers who will steer the right questions to ask the machine as well as assisting computers stuck in a “can’t decide” situation. Ultimately AI will make processes more efficient and error-free, but it needs to be complemented by experts with years of experience.
Satellite Evolution - NewSpace International - May/June 2021