CNN's Tomorrow's Hero profile Meet Amber Yang. She's trying to prevent a space debris catastrophe describes an investigation into the use of a convolutional neural network (CNN) to predict possible collisions in low Earth orbit.

If I understand correctly, searches for and predictions of potential collisions (such as SOCRATES for example) are done by raw number crunching, propagation of all known objects with something like SGP4 combined with some basic algorithms that only scrutinize pairs of objects with some possibility of intersecting orbits. One might not compute distance of closest approach between a pair of circular orbits if one were GEO and the other MEO for example.

How might one apply a CNN to this problem? Plausible speculation is welcome, as this is a somewhat new approach.

edit: I've found a short but interesting introduction to the problem (see page 6 for a flow chart), but I don't think it's meant to be a complete description of the technique.

below: plot from the CNN article.

enter image description here

Good video, if you are short of time, watch the last two minutes at least!

  • 2
    $\begingroup$ Although it's a space question it's more of a machine learning/AI questions so much more applicable to Data Science.SE. $\endgroup$
    – GdD
    Commented Jan 16, 2018 at 12:53
  • 1
    $\begingroup$ @GdD I know what you mean, but If I tried to ask this there, I'd probably have to explain orbital mechanics. Here there are many users who understand how orbits work, what parameters are available and how the relationships between parameters relate to the possibility of intersection between orbits, maybe even be people familliar with other applications of AI to orbits. I'm not sure how to think through and then explain all of that in a question there. If there is some receptiveness to answering this question, in its current or a similar form, there I'm certainly flexible about moving it. $\endgroup$
    – uhoh
    Commented Jan 16, 2018 at 13:12
  • 1
    $\begingroup$ @Chris I've added both a link to a pdf, and a note that the graphic is from the CNN article. $\endgroup$
    – uhoh
    Commented Jan 16, 2018 at 13:59
  • 1
    $\begingroup$ I'm not sure how many physicists we have over on AI, but we certainly have a few Neural Network experts. Even without the formal physics background, they'd probably be able to give insight on NN approaches to this and similar problems. Not to mention, it sound like a problem with a lot of game theory aspects. $\endgroup$
    – DukeZhou
    Commented Mar 28, 2018 at 21:25
  • 1
    $\begingroup$ Here's a more recent article about Amber Yang; and her company's website. $\endgroup$
    – Chris
    Commented Apr 26, 2018 at 3:06

1 Answer 1


As a data scientist with an interest in space stuff, I am well positioned to give an answer this question. I must confess I don't understand fully what is going on in this system from just the lecture slides, so my answer will be more of an educated guess.

At first glance it doesn't really make sense to use a neural network here. The sort of problems that neural networks are amenable to solve tend to be rather soft and fuzzy (both in the mathematical sense and the is-this-picture-a-cat? sense). Physics simulation problems have well-defined rules, and the best way of solving them is usually through good old fashioned brute-force simulation.

What I think is happening here is that the neural net training is being used to create a statistical object, a sort of probability distribution. One way of thinking about neural networks, is that they're acting as a form of data compression. By training a neural net on something, you are essentially fitting a complicated multidimensional curve made of hundreds of parameters onto data that could contain millions of points. (In a way, humans do this too. We take a kaleidoscopic variety of individual experiences and make sense of them through a relatively small number of rules of thumb, which we then use to guide decisions in future events that are unlikely to be exactly the same as a previous event).

As mentioned in LeWavite's answer to one of your previous questions, for n objects you have 1/2 n(n-1) possible collisions to worry about. There's about ~17000 bits of space debris that are currently tracked, meaning there's ~ 144,500,000 possible collisions to sort through, a rather unwieldy number.

The key slide from her presentation is this one: Picture of two different neural nets (You'll probably need to enlarge it to see better, the original wasn't that good)

From what I can gather from this diagram, there are two neural networks with two different functions. The diagram on the left looks to be some kind of control systems state diagram of the type roboticists use. The red neural net takes in 5 orbital parameters (the ones needed to define the orbit, but not where the object is in the orbit). It's not clear to me what the three output parameters (Y(1),Y(2),Y(3)) are, they don't seem to appear anywhere else.

The blue neural network appears to act in a similar fashion to an extended Kalman filter, in the sense that the predictions of the red neural network are being continually updated by new data of the tracked objects as their orbits naturally shift from their keplerian ideals. By training the neural network on these updates, it sort of learns a 'sense' of how the object's orbits tend to change over time.

That's about all I can extract from the diagram. Considering this is a proprietary system, I doubt there'll be much more than that.

I do question the usefulness of using convolutional neural nets. They are used for situations where the individual columns of data are related spatially - like points on a grid. You mainly see them with image-processing neural nets, but they can do other stuff, like generating terrain from drawn lines. There are only 5 input parameters in both of the neural networks shown in the talk, and they represent different concepts, so I fail to see how CNNs can help here.

  • 1
    $\begingroup$ I felt the same way; it seems like a mismatch, thus the question. Start-ups can take several twists and turns as they grow, so it will be interesting to see what happens. Thanks for your answer! $\endgroup$
    – uhoh
    Commented Feb 15, 2019 at 15:43
  • 1
    $\begingroup$ I just saw this answer to Is the classical three-body problem solvable? $\endgroup$
    – uhoh
    Commented Jul 1, 2020 at 3:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.