- The new neural network model improves the accuracy of laser-ranging telescopes.
- This allows scientists to pinpoint the position of smaller debris in Earth orbit, without increasing the sensitivity of telescopes.
Space debris is defunct human-made objects in Earth orbit that are no longer useful. They are generated during space activities, and mainly come from the final stage of carrier rockets, and unwanted materials of spacecraft disintegrating in the orbit.
As of 2019, there were approximately 20,000 artificial objects in orbit above the Earth, including 2,218 operational satellites. However, these are the only objects large enough to be traced.
More than 130 million pieces of debris are smaller than 1 centimeter, about 1 million pieces are 1-10 centimeters, and over 30,000 pieces are larger than 10 centimeters. They all are in orbit around the Earth and their existence seriously affects the safety of a spacecraft.
At present, there are more than 50 global laser observation stations that monitor space debris. However, tracking them is an extremely challenging task: the smaller the object, the more difficult it is to detect and track.
A new study presents a neural network model that improves the accuracy of laser telescopes, allowing scientists to pinpoint the position of smaller debris.
How Accurate It Is?
To pinpoint orbital debris, scientists utilize a method called laser imaging. This involves beaming high energy lasers into space and using a telescope to pick up the signals that are reflected back from orbiting debris. These reflected signals are then used to evaluate how far debris is. The process is similar to how bats use echolocation to track prey.
But since the smaller debris does not reflect much light, it is difficult to locate them precisely. Although previous techniques have enhanced laser ranging detection of debris, they can pinpoint a piece of debris only to a 1-km level.
The new method, on the other hand, is capable of finding a piece of debris as small as 1 square meter in size that’s about 1,500 km away.
To achieve this, researches used a backpropagation neural network, which is optimized via two correcting algorithms: the Genetic Algorithm and Levenberg–Marquardt.
An illustration of space debris as could be seen from high Earth orbit | Wikipedia
The neural network helped telescopes stabilize their pointing abilities and recognize weak signals of small pieces of space junk. This is all done without increasing telescopes’ sensitivity or making any hardware upgrades.
Also, telescopes can now detect debris more accurately in localized regions of space without producing many false positives.
Researchers tested their method against three standard models: the mount model, the basic parameter model, and the spherical harmonic functions model. They found that the accuracy of the neural network is superior to these three standard models, and it also overcomes the shortcomings of slow convergent speed.
To demonstrate the capability of the neural network, researchers used observational data of 95 stars to solve the algorithm coefficient from all 4 models. They assessed the accuracy of detecting 22 other stars. Not only the new algorithm proved to be most accurate, but it could also be easily operated with decent real-time performance.
The team plans to further optimize the neural network to spot even smaller debris.