2

Introduction

To kick things off, I'd like to introduce myself as a 23-year-old novice learner when it comes to space exploration but I've been listening to astronomy-specific podcasts since 2022 and I've been trying to learn as much as I can about the subject since then despite this being kind of irrelevant to my field of study at university (CS).

Overview

Overall, there is basically a series of mathematical and computational processes involved with which the spacecraft is directed to its destination. I'm going to make some assumptions for the rest of the overview. For instance, I know just a little about the Hohmann transfer orbit which is a way for spacecraft to gradually align their path with the orbit of their target object. This, however, is calculated fairly accurately by a computer that is either on the spacecraft itself so that it can be somewhat autonomous as it approaches the target(but I'm not quite sure if that's how missions like Voyagers or Pioneers were done) or operators manually or in an automated fashion, send the individual instructions required to direct the spacecraft to the right path, to the spacecraft using earth-based radio transmitters.

The Crux

What methods have been used for guiding JUICE and how technologies such as AI, CNN-based Image Processing, and optical navigation have been used to increase efficiency?

Navid
  • 23
  • 5
  • Please edit the question to limit it to a specific problem with enough detail to identify an adequate answer. – Community Sep 14 '23 at 04:22
  • I think this question is very good, but maybe request an edit from a member of the community who's been here for a while. It's, well, LONG. Unnecissarily. And I don't mean that in a bad way, it just rambles, and I'm guilty of that too. Maybe ask about one specific mission: "How does JUICE navigate autonomously?" Or ask about an agency: "What is the navigational-technology preference for ESA?" – AnarchoEngineer Sep 14 '23 at 04:26
  • @AntiElonGuy Thanks for your suggestion, I agree that it was too generic but I think after the change I made to the title, I should be fine, I suppose. – Navid Sep 14 '23 at 04:34
  • 1
    Yeah, just a suggestion! Welcome to Stack Exchange! – AnarchoEngineer Sep 14 '23 at 05:18
  • 2
    Related question: https://space.stackexchange.com/questions/5845/how-do-probes-locate-themselves – Hobbes Sep 14 '23 at 07:48
  • 2
    and https://space.stackexchange.com/questions/942/how-can-spacecraft-navigate-without-contact-from-earth – Hobbes Sep 14 '23 at 07:51
  • No spacecraft navigates fully autonomously. They are always in communication with Earth, recieving orders and sending data on a daily-to-weekly basis and sometimes system updates. – AtmosphericPrisonEscape Sep 14 '23 at 10:09
  • @AtmosphericPrisonEscape That's what I intuitively thought, however, what I'm interested to know is whether they are capable of being fully autonomous during Orbital maneuvers or not. As a CS student, I also want to know how in this context, Machine Learning techniques can actually accelerate navigation. – Navid Sep 14 '23 at 10:33
  • Answering this would take a book "How many methods are there for guiding spacecraft in space and how technologies such as AI, CNN-based Image Processing, and optical navigation are used in space exploration missions today to increase efficiency?" If you edited your question to take out all the extraneous bio and fluff, and just asked about JUICE, it could be OK. – Organic Marble Sep 14 '23 at 12:49
  • 1
    Remember that interplanetary navigation was quite successful in the 1960's and '70s (Viking mission had Mars orbiters and landers) when the on-board capabilities were tiny. Improvements to interplanetary navigation will likely come from improvements to propulsion rather than any refinements to control systems, on-board or otherwise. – antlersoft Sep 14 '23 at 13:50
  • 2
    @OrganicMarble I see your point, just made it more specific to JUICE – Navid Sep 14 '23 at 18:01
  • 1
    Close vote retracted after edit narrowing focus to JUICE. – Organic Marble Sep 14 '23 at 18:04
  • @antlersoft I don't want to pretend that I've read enough articles on propulsion but I haphazardly know that nuclear thermal and nuclear-electric propulsion are under research. Nevertheless, I really want to know about areas where ML and interplanetary missions such as JUICE intersect, and for that reason I thought the best place to start was the navigation system. – Navid Sep 14 '23 at 18:17
  • 1
    @antlersoft I can imagine a time in the future when probes like New Horizons explore Oort cloud objects and the concern is unknown companions and debris fields where the ability to detect/measure them locally, send info to Earth, calculate a new, safer flyby trajectory and send it back to the probe is not possible. At that point some refinements to control systems on-board may be of value. – uhoh Sep 15 '23 at 00:27

1 Answers1

3

It would appear 'Autonomous Navigation' may be generous in terms of what JUICE is doing. This Airbus document limits itself to describing edge detection to fine tune instrument pointing on the fly, and smart upset recovery. This is useful for maximising science collection and reducing risk but does not make the craft able to self direct around Jupiter on long time scales.

All space probes including JUICE are severely power and mass constrained, and JUICE in particular has to operate in a highly hostile radiation environment meaning that it cannot fly serious computer power. In the case of Juice the solar panels have a designed power output of 850 watts (enough for a gaming laptop but not server farm).

Navigation in terms of current position and velocity is a complex question. Orientation can be determined with star trackers since the early ballistic missiles of the 50s, and most space craft since the voyagers have been able to autonomously work out which way they are facing and then find earth should communications fail. However getting position/velocity in absence of GPS type beacons is much harder and is normally done by measuring Doppler in the downlink to earth. JUICE does have a low power radar that may be aiding flybys by confirming distances are as expected though this is not explicitly stated.

This means that the bulk of the trajectory determination and future planning will still be happening on earth, with the 'Autonomous' allowing minor errors in sensor pointing or software to not result in a missed opportunity to collect data or loss of vehicle.

True AI has little place in space exploration since by definition it will behave unexpectedly, which is hard to justify on billion dollar missions. Less bleeding edge systems such as image processing algorithms to support navigation and identify boulders during touch down certainly are used. While not AI the posited cause of Chandrayaan 2's failure is an example both where some degree of autonomy might have saved the mission (having 'plan has failed, just land somewhere' code), but also a situation that would trap a poorly designed AI (ending up inverted).

True autonomy will probably not show up in deep space missions until there are enough duplicate missions in operation that failures can be allowed for and accepted while the system learns.

GremlinWranger
  • 22,391
  • 1
  • 56
  • 87
  • 3
    "True AI has little place in space exploration since by definition it will behave unexpectedly..." e.g. "I can't do that, Dave." – Organic Marble Sep 15 '23 at 11:58
  • @OrganicMarble To play the "fun at parties" guy - "true AI" is usually a synonym for "general AI" in contrast to "domain AI" which is a synonym of "machine learning" and that is a synonym of "weird math". And - to my knowledge - that it has not been developed yet. Although we are preeeeety close. Scary times. – Vorac Sep 24 '23 at 08:38