Tracking Dead Satellites With Optics and Light-Curve AI

Space situational awareness (SSA) centers monitor defunct satellites to prevent orbital collisions and plan future debris removal missions. Since dead satellites no longer transmit telemetry data, operators rely on ground-based optical telescopes and advanced artificial intelligence to figure out how these objects are behaving. By combining optical streak detection with non-resolved light-curve AI analysis, SSA centers can accurately determine the spin rate, orientation, and overall attitude stability of tumbling space junk from thousands of miles away.

Optical streak detection is the first step in this characterization process. Ground-based telescopes take long-exposure photographs of the night sky. Typically, a telescope tracks the slow movement of background stars, which causes a fast-moving defunct satellite to appear as a long line, or "streak," across the image. Alternatively, if the telescope tracks the satellite itself, the background stars become streaks. Automated software detects these streaks and measures their exact pixel brightness from start to finish. Instead of being a solid, uniform line of light, a streak often looks like a dashed line or a wave of changing brightness. These fluctuations happen because the satellite is tumbling, causing sunlight to glint off different surfaces—like flat solar panels or curved fuel tanks—as it moves.

Because most satellites are too far away and too small for a telescope to capture a clear, detailed picture of their shape, scientists refer to them as "non-resolved" objects. Even the best telescopes only see them as single points or thin lines of light. To bypass this physical limitation, SSA centers extract the changing brightness data from the optical streak and plot it on a graph over time. This graph is called a light curve. A light curve acts as an optical fingerprint for the satellite's rotation. A stable satellite exhibits a smooth, predictable light curve, while an unstable, chaotically tumbling satellite produces a jagged, irregular pattern.

This is where rotational artificial intelligence steps in to decode the light curves. Historically, human analysts had to manually compare light curves against physics-based computer models, a slow process that cannot keep up with the thousands of defunct satellites currently in orbit. Today, SSA centers use machine learning algorithms, such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), to automate this task.

Engineers train these AI models using millions of simulated light curves alongside historical data from satellites with known behaviors. The AI learns to recognize how specific satellite shapes, materials, and spin states reflect sunlight at different angles. When a telescope feeds a new streak light curve into the system, the AI analyzes the frequency and amplitude of the brightness peaks in seconds.

Through this analysis, the AI characterizes the satellite's attitude stability. It classifies the object into specific rotational states, such as remains three-axis stabilized (holding a steady orientation), spinning uniformly around a single axis, or tumbling chaotically out of control. Knowing this stability is critical. If an SSA center knows a defunct satellite is tumbling fast, they can better predict how solar radiation pressure and atmospheric drag will alter its orbit, ultimately keeping active, operational satellites safe from unexpected collisions.