Is Satcom Ready for AI?

By: Helen Weedon

AI has been a big talking point at every satcom conference and industry event over the last few years and there’s little doubt that the technology holds great potential for the industry. While there’s a lot of talk about what AI can theoretically do, it’s fair to ask “How much progress is the industry making in moving from hypothesis to reality?” And, “Is the satcom industry behind or ahead of other industries in terms of adoption?” Members of the Satcom Innovations Group (SIG) recently shared their views on where the biggest potential for AI applications lies in satcom operations and its current state of adoption. 

Where can AI bring most value?

Automation is probably one of the first things that comes to mind - with AI being used in areas where it will improve efficiency and reduce human error. Tasks that are repetitive, highly complex, or involve frequent decision-making tend to offer the clearest opportunities for efficiency gains and error reduction. As Alexander Jeuck, N@SAT SINGAPORE, says, “in principle, it should help to make satcom more reliable and more affordable, and therefore also more competitive”. 

Spectrum management is another obvious area where AI can make a difference. As Erwin Greilinger, Eviden, explains “AI can enable effective usage of available RF spectrum. In certain areas spectrum is a scarce resource and an optimised frequency planning and spectrum utilisation is needed. AI could dramatically improve the optimal usage of the available spectrum.”

Jörg Rockstroh, Work Microwave, adds: “I’m convinced that all ‘big data situations’ offer a fantastic field in our industry where AI brings significant advancements. Be it physical channel behaviour, link availability models and predictions, network/traffic load (re-)distribution, everything that is defined by complex influences and where simplified models lose on efficiency. AI, and in particular agentic AI, is a next level tool to improve handling of these situations compared to current model-based coping mechanisms.”

While Bob Potter, Global Invacom, sees a lot of opportunity around using AI to better analyse signals to detect interference. Rather than using conventional methods involving demodulation and signal cancelling, which involves looking inside signals to see if there is co-channel interference, Bob says “AI could be trained to analyse signals better, through carrying out quicker or more intelligent analysis based on its own learnings from signal shapes”. 

Although Najeeb Khan, Serana, agrees that “signal characterisation is one of the prime applications that comes to mind when thinking of AI in satcom,” he sees AI as having a fundamental role above and beyond the RF path. In his view, AI will bring most value at network and operations level where it can effectively monitor and understand what is happening across a network and then decide what action to take: “The real opportunity for AI lies where humans are the bottlenecks. For example, traffic steering, where we have a self-healing network, where the network learns the normal behaviour, spots degradation and takes some sort of action, like rerouting the traffic or steering to a different satellite or orbit.” 

Khan went on to explain what this could mean in practice: “If you have a payload failure at two a.m. at night, that’s usually a war room situation. But if AI could help find what the payload loading looks like, what the gateway capacity is, what the interference is, and take SLAs and contracts into account, reason about it and come up with a plan, that’s the sort of application AI enables and really brings operations to the next level.”

From the operator perspective, there is potential for AI to turn monitoring from a reactive process into a proactive one. For example, if there is a failure or interference on a transponder, an alert can be sent when levels are decreasing or a combination of different parameters. This provides advanced warning when an instance occurs, and the operator address it before customers need to be moved to other transponders.