AI-Controlled Satellite Achieves Real-Time Fault Detection

As satellites become increasingly critical to global communications, navigation, weather forecasting, and defence, the pressure on orbital infrastructure has never been greater. Yet, despite their sophistication, most satellites remain dependent on ground-based teams for decision-making, leaving them vulnerable in an environment where every millisecond counts.

Now, a research team at UC Davis has unveiled a breakthrough: the first satellite equipped with onboard AI capable of real-time self-monitoring and predictive diagnostics. Developed in collaboration with Proteus Space, this innovation could mark a turning point in satellite autonomy.

What challenges do current satellites face, how does this AI-driven approach work, and what could this mean for the future of space operations?

The Challenge With Satellites – No Autonomous Control

Satellites have fundamentally reshaped human civilisation. From GPS navigation and weather forecasting to global communications and Earth observation, they’ve become a backbone of the modern world. We rely on them more than most people realise, and the system works impressively well considering just how fragile it really is.

But here’s the catch: space is getting crowded. With thousands of active satellites and an alarming amount of debris, dead satellites, rocket stages, and even flecks of paint, low Earth orbit is starting to resemble a high-speed junkyard. And when you’ve got that much metal flying around at 28,000 km/h, things get risky fast.

One of the biggest headaches? Collision avoidance.

Believe it or not, most satellites don’t have any kind of onboard autonomy for handling this. Instead, when two objects are projected to come uncomfortably close, and “close” in space can mean within a few kilometres, engineers on the ground have to step in. They manually assess the risk, calculate the odds of collision, and then issue corrective manoeuvres to the satellite. That means everything from tracking potential threats to planning and uploading commands is handled by teams on Earth, usually under tight time pressure.

There are a few problems with this manual approach. First, latency. Space is big, but network lag still matters. When every second counts, delays in communication or decision-making can be the difference between a successful dodge and a million-dollar pile of orbital confetti. Second, prediction. You need accurate and timely data on every nearby object, which is easier said than done given the chaotic nature of orbital debris. And finally, your command has to reach the satellite, if it’s on the wrong side of the planet, or your uplink is down, you’re out of luck.

All of this highlights a core limitation in today’s satellite infrastructure: the lack of autonomous onboard control. Despite being multi-million-dollar machines operating in one of the most hostile environments imaginable, most satellites still wait patiently for someone on the ground to tell them what to do. As we launch more, including mega-constellations with thousands of units, this manual, reactive model just won’t cut it.

AI-Powered Satellite to Self-Monitor in Orbit

A UC Davis research team has developed the first satellite with onboard AI capable of real-time self-monitoring and predictive diagnostics. Built in collaboration with Proteus Space, the mission is slated to launch in October 2025 from Vandenberg Space Force Base, just 13 months after project initiation. That compressed timeline alone demonstrates how far satellite design and prototyping methods have come.

The satellite’s core innovation is a digital twin embedded directly into the spacecraft. Unlike conventional ground-based models, this twin runs onboard, using AI to monitor battery health, analyse voltage trends, and predict performance. It reduces the need for constant communication with ground control, enabling autonomous operation and early fault detection.

“The spacecraft itself can let us know how it’s doing,” said Adam Zufall, graduate researcher on the project.

This capability is powered by embedded sensors and machine-learning algorithms that adapt over time, improving predictive accuracy with each orbital cycle. A secondary payload has also been included to validate the system’s predictions under real orbital conditions.

Traditionally, satellite development spans several years. This mission, however, leveraged a rapid-prototyping approach that cut design-to-deployment to just over a year. The satellite bus was supplied by Proteus Space, while UC Davis researchers focused on the custom payload that drives the AI-based monitoring system.

The spacecraft will operate in low Earth orbit for up to 12 months. After deactivation, it will remain in orbit for about two years before safely burning up in the atmosphere, complying with orbital-debris mitigation rules. In addition to improving safety, this also sets an example for sustainable space operations.

Onboard AI and real-time diagnostics shift mission autonomy from ground teams to spacecraft themselves. This reduces operational costs, increases system resilience, and enables more frequent, faster-deploying missions. UC Davis researchers view this satellite as a proving ground for smarter, self-aware spacecraft. If successful, it sets a new precedent for agile, intelligent space systems.

How AI-Powered Autonomy Could Transform Satellites

Giving satellites the power to make their own decisions is a game-changer. Right now, collision avoidance is a headache that requires constant ground intervention, complicated predictions, and a lot of uncertainty. With onboard AI, that problem becomes simpler. Instead of predicting every possible future collision days in advance, satellites could evaluate risks in the moment, deciding when and how to act, making real-time course corrections as situations evolve.

Space is chaotic. Not every threat is foreseeable: debris can appear unexpectedly, or an untracked object might suddenly cross paths. With current ground-dependent systems, rapid, unpredictable events often leave no time to respond. AI-enabled satellites could cut through the uncertainty and act instantly.

Of course, autonomy in orbit isn’t something you deploy without caution. Safety is paramount. A wrong decision, an unnecessary manoeuvre or a failure to avoid debris could be catastrophic, creating more debris or damaging the satellite itself. The algorithms and hardware must be rock-solid, extensively tested, and built with fail-safes.

If the UC Davis prototype proves effective and reliable, it could set a new industry standard. Future satellites might all ship with self-monitoring and decision-making capabilities as standard, shifting control from slow, human-in-the-loop processes to fast, autonomous operations. In short, this tech won’t just improve satellite safety and efficiency; it could reshape how we manage the increasingly crowded skies above us.