Martin Halliwell is a Partner at NewSpace Capital, one of the world’s first private equity firms devoted exclusively to growth-stage companies working in the space technology sector. He formerly served as Chief Technology Officer of SES, where he led global technology and R&D from 2011 to 2019. Halliwell contributed this article to Space.com’s Expert Voices section.
There are now some 16,000 active satellites in orbit. Market intelligence firms like Novaspace say that up to 43,000 satellites will be built and launched over the coming decade. Others, such as Goldman Sachs, think as many as 70,000 new low-Earth orbit (LEO) satellites could be launched in the next five years alone. The exact number is less important than what these predictions reveal: that satellite constellations are getting bigger.
This is a good thing. Space is a crucial enabling force in the modern world, touching just about every part of ordinary life. But as these constellations grow, it will become increasingly difficult for human operators to run them. A network of a few satellites can be managed from the ground; a network of hundreds, or even thousands, cannot be managed in the same way. These satellites must share data, avoid interfering with one another, meet changes in customer demand, and deal with technical problems. Some decisions must be made in seconds, without waiting for instructions from Earth. This is why onboard processing, automation and artificial intelligence are becoming all-important. Satellites will need to sift through and analyze more information in orbit, then act within clear limits set by human operators.
In-orbit processing
One major use for AI will be processing data in orbit. At present, most satellite data is sent back to Earth before it is cleaned, sorted and analyzed. Earth-observation companies, for example, may collect huge amounts of imagery to track weather, crops, disasters, land use or emissions. They must then process all of it on the ground before they can give customers useful information. If satellites can do more of this work in orbit, they can send back only the data that matters. This would cut costs and save time. In defense, where speed can decide the outcome, it could help people act much faster. Onboard processing could also reduce reliance on ground systems, which may themselves be attacked or disrupted. Satellites remain vulnerable, but moving some analysis into orbit could make the wider system more resilient.
Moving capacity
Satellites must be able to spot patterns, respond to changes and adjust their plans in near real time. This makes them well suited to AI. One important use is managing network capacity. Demand for satellite bandwidth is always changing. It may rise over cities at busy times, around major events, in disaster zones, on aircraft and ships or during military operations. AI could track these changes and decide where a satellite’s beams should point, how much power each beam should use and when capacity should be moved elsewhere. This would help satellite networks send more capacity to places where it is needed most, instead of sending the same amount to the same places all the time. The result would be a more efficient use of the satellite’s limited power and available spectrum.
What if AI gets it wrong?
The barriers are not only technical. They are also legal and organizational. Most laws, contracts, insurance policies and operating rules were written for systems controlled by people. They assume that a named person makes each important decision. AI makes responsibility less clear. If an AI system points a beam at the wrong place, disrupts another service or causes damage, it may be difficult to decide who is at fault: the satellite operator, the manufacturer or the software company. Operators must also decide how much control they are willing to give to AI. Many are happy for AI to offer advice. Fewer are ready to let it change a network or move capacity without human approval. For this reason, adoption is likely to be gradual. Operators will need clear rules about which decisions AI can make by itself and which must still be approved by a person.
Spacecraft design and manufacturing
AI will affect more than the way satellites are operated. It could also help engineers design and build them more quickly. Engineers can use AI to write code, search technical documents, test ideas and suggest solutions to design problems. It could produce early versions of spacecraft structures, antennas, power systems, payloads or mission plans. Human engineers would still need to check, test and improve the work, but they would not always have to start from scratch. This could shorten development times, cut costs and allow smaller teams to do work that once required much larger organizations. AI could also help in factories by spotting faults, predicting delays and showing when equipment needs maintenance. It will not replace skilled engineers. Its main value will be in doing routine work quickly, so people can spend more time on difficult decisions, safety and the hardest technical problems.
Building trust
To unlock AI’s full value, the industry will need secure, affordable environments in which organizations can train and deploy models without losing control of information. These environments need strong cyber security, clear access rules and ways to keep each organization’s data separate. Some models may need training inside company or government networks so sensitive data never leaves. Technical safeguards will help, but trust matters, too. The industry will also need better standards, clearer legal responsibility and reliable testing. The technology itself is advancing quickly. The harder task is creating the rules, safeguards and confidence needed to use it. AI may soon be ready to run more of a satellite network. The question is how quickly the industry will be ready to let it.
