We send an unmanned probe to Alpha Centauri. How do we control it when it's there?

Factual Questions
21

I guess the problem is - how easy is it to identify planets remotely? Right now, we identify them by two means- wobble and transit.

We measure the wobble because technically the star and planet orbit the common center of gravity, which implies it better be one heavy planet to create an appreciable wobble in a star. (There’s a secondary version of that where we see the doppler shift in the light as the star moes around that common center of mass). Considering that for a pretty much balanced system like earth-moon, that common center of gravity is still deep below the earth’s surface, star wobbles are pretty miniscule.

Transit - it’s amazing that exists, except for super-jovians orbiting red dwarfs. The amount a sun dims as a planet transits is amazingly difficult to detect matched only by the odds that we would be inline with a transiting planet in the first place.

But in-close planets are probably easier to detect from a few light-months away since being close to the star, they would be fairly bright (on the opposite side of the sun) and moving fast and have a short period. And typically, planets orbit in roughly the same plane as the star’s rotation, so a clue where to start looking.

The other thing to understand is the magnitude of the challenge. IIRC the solar escape velocity is about 17mi/sec. An 80-year voyage, 0.05c is on the order of 9,000mi/sec. This is not something that will be accomplished with chemical rockets.

22

The odds aren’t all THAT long against transits. For an Earthlike planet orbiting a Sunlike star, there’s on the order of a 1% chance that they’ll be lined up well enough to transit. And there are far more than hundreds of candidate stars, so we’d expect to find plenty. Even more importantly, we can calculate precisely how likely such a lineup is, for any given star size, planet size, and orbital radius, meaning that we can confidently extrapolate about the ones we’re not detecting, based on the ones we are detecting.

Now, that’s still not much help when we’re looking at just alpha Centauri, or at a short list of stars within, say, 10 ly of Earth. But as scr4 says, there are new technologies within our horizon that could, in fact, detect extrasolar planets directly. And alpha Centauri and the other nearby stars would probably be among the first places we look, once we built such a telescope.

23

Of course, the algorithms would give weight to size (gravity), position near the goldilocks zone for heating, spectroscope of atmosphere composition, etc. Presumably it would include detachable sub-probes to check alternative targets (Although bombarding a habited planet with probes going 10,000 miles a second might not be the best idea.) The probe would have a decent amount of time to analyze observations on approach to pick the best targets, and we could use earth observations to add guidance factors.

Also keep in mind that unless the probe does brake - even at 1/20th c, it would cross the diameter of earth’s orbit 16 light-minutes - in 320 minutes or just over 5 hours. the diameter of Jupiter’s orbit in 50 hours or 2 days. You think the Pluto flyby was a long wait for a short look, imagine 80 years to get maybe a 1-hour closeup of the target planet. I hope by then longer-distance telescopes work well.

24

If it is just a fly-by, I hope we could make it a large cigar-shaped object, with ‘Oumuamua painted on its side. :smiley:

25

What happens if we smash it into a planet at .05 c? Tell the locals the neighbors say “hi!”

26

And therefore the onboard computer can do any emergent calculations as pertubations from the expected trajectory, making its job a lot easier.

27

We can do it with “artificial intelligence” without resorting to decision trees. Though honestly the way we’d do it today is barely an upgrade from that.

The purpose of your probe is to maximize total expected reward, with shorter term reward better than longer because the probe has a finite lifespan.

What is “reward”? Obviously detailed information about planetary objects of interest to humans. Where detail can be defined as predicted resolution of pictures (closer you get, the higher the resolution) times the number of them captured, and also humans are interested in all the planets, sorted in order.

Present day programmers would then choose an architecture, probably some flavor of neural network, that would actually produce the high level command decisions that would plan out where the probe intends to fly and where to point it’s instruments. We would then test the candidate architectures against simulations of what might happen during the mission and select the architecture that performs the best, probably using some heuristic to measure how broad-scope a variety of situations it handles. Note that the simulations will randomize parameters like the laws of physics a little and inject things like sensor noise and failed instruments in, so that the final policy network is more robust.

Obviously this won’t be present day. It isn’t magic that is required, it’s an absurdly massive starship loaded with fusion fuel. (probably hydrogen slush and boron/lithium, or antimatter). Millions of tons of fuel, all in orbit. And fusion engines that work on small quantities of gas and burn continually, versus the fusion devices we have built that only function as bombs.

This would take decades at a minimum, at the present rate of progress it won’t happen this century. (well, technically the rate of progress seems to be accelerating and obviously if you take the trend to the logical limits it would happen before 2050, but the Singularity is a controversial topic…)

A more practical probe would have vastly more sophisticated AI. It would control dozens of robotic repair drones and have an onboard factory, using nanotechnology, able to recycle any component on board back to a new component. It would use a far more robust algorithm that is capable of handling any situation where the (1) problem is solvable with the information the probe has, which would include deep technical databases about most human knowledge (2) where the problem can be solved with the tools the probe has onboard and the time and resource constraints, etc.

Superhumanly competent at fixing and operating itself, essentially.

But we can only really speculate on how to build such a thing. Present AI papers show that this is probably eventually possible.

28

If I recall my high school physics - E=(1/2)Mv^2

So assume a small probe, only say 10 tons or about 10,000kg travelling at 15,000,000m/s - and that energy is translated into heat or kinetic impact within the 100km of an earth atmosphere. It’s one way to make sure future colonies don’t have to deal with dinosaurs.

29

But bear in mind that each Kg of fuel you burn and exhaust is one Kg less mass you have to push, so your effective thrust increases as you move forward because it is working against less mass.

30

If you read something that says we measure the star’s wobble to detect planets, they mean we measure it by doppler shift. And yes, it is subtle, but astronomers are apparently good at seeing subtle. And both this method and transits are best at finding large planets in close orbits, so those planets are way over-represented in the list of known exoplanets.

There was an attempt by someone in the last century to measure planet-induced wobble by the positions of a star as it moved across the sky. He claimed to have discovered a couple planets at Barnard’s Star, but it turned out to be a systematic error.

Oh vell, back to zee drawink board…

31

A not-insignificant part of the challenge is to be able to design a machine that will remain functional after decades of fast travel through interstellar space - it is ‘just’ an engineering problem, but it’s a tough one - even with our interplanetary probes, which are engineered to a superbly high standard, systems fail or misperform in the few short months that it takes to get where they’re going.

32

Yes, but it’s different sorts of bias for the two methods. The wobble method can’t detect small, far-out planets at all, so if that were all we had, we would have no idea how many such planets we were missing. The transit method can, just at lower (but known) rates, so we can tell how many we’re missing.

Simulations of how solar systems form are a good example of how bias can creep into scientific thought. Back before any exoplanets were discovered, the only solar system we knew of was our own, and so when people created simulations of how solar systems “should” form, the simulations usually produced small, rocky worlds relatively close to the star, and large, gassy worlds further out, just like ours. If anyone came up with models that didn’t do that, they’d have gone under extra scrutiny and eventually been rejected as unrealistic.

But then we started finding exoplanets, and at first, all we found was hot giants: Planets closer to their star than Mercury, and larger than Jupiter. That wasn’t really much of a surprise, because that’s all that we could discover: The only new information was the mere fact that such planets were possible. And yet, people started revising their simulations, and now found that the vast majority of simulated solar systems had hot superjupiters, and very few had inner rocky planets and outer giants, in the same proportions as the ones we knew about (even though everyone knew that the ones we knew about were biased).

Then the transit method kicked in, and it turns out that solar systems like ours aren’t so rare after all. And so now, that’s what the models mostly show, again.

33

FWIW, at .05c you’d cross the diameter of Neptune’s orbit in ~83 hours. The time spent in the vicinity of a planet taking pics would be measured in minutes. Precise camera rotation to get a steady pic would be very, very difficult and result in large delays between pics. Would be lucky to get a sequence off at the level of Mariner 4. And you’d have probably a single planet flyby.

To slow it down from that speed means tremendous increased weight at that end which in turn mean stupendous increase in weight at the start.

34

You write as though decision trees and artificial intelligence are mutually exclusive, but they are not. An autonomously acting computer that takes signals from its environment and chooses an appropriate course of action with the help of a decision tree is artificially intelligent, and this is true whether that tree is hard-coded or automatically learned. Decision trees might not be the latest and greatest thing in the AI toolbox, but they’re among the most common.

35

It would seem like the probe must be able to compensate for malfunctions, live in a broken state and make it work. We have robots that can do this, but that is mechanical systems which it relearns to use and move itself after parts breaking. We would have to have the probe do this electrically.

As for getting into orbit, or just slowing down, it would seem like that ability would just would make the trip there much longer. Perhaps a solar sail could be used, which may be part of a transmitter and solar (steller) power as well. The only other thing that could reasonably stop the probe from leaving the system would be a to use a planet, perhaps a aerocapture maneuver, but traveling through a atmosphere at 0.25c wouldn’t leave that much of the probe left, and what is left would no longer be considered a probe.

36

I assume you are talking about something like Breakthrough Starshot (I haven’t read any of the follow on answers so this is probably redundant). If so, yes, it would just be a flyby. They don’t plan for it to maneuver in the Alpha Centauri star system, just fly by at a fraction of the speed of light. It would have instruments on board (though this thing is seriously small) to observe things like planets, but really it’s more a proof of concept that could get us there in, IIRC, something like 20-40 years (it will take over 4 years just for the data, whatever it is, to get back). It won’t brake at all…it will basically just fly through then out of that solar system and, in theory, out of the galaxy until and unless it hits something.

37

In order to make it worth while to send a probe to Alpha Centari we need to accomplish several things

  1. Make a ship that travels fast enough such that a significant faster ship won’t be invented during the transit time. (There is no point in starting a 500 year journey if in 200 years technology will improve enough to cut the journey time in half.)
  2. Will survive potential ultra high speed collisions on the way there
  3. have a way to send a signal back that can be received and understood back on earth
  4. have a self automated control system to guide its journey.

Of those 4 problems, the last one is the easiest and probably the only one that can be solved with currently available technology.

38

Well, the ones that are actually part of the proposed mission (meaning they are within our capabilities) are 1 and 3. 2 is impossible…the thing will be going at relativistic speeds. If it hits anything it’s going to blow up like a big bear. I’m not sure about 4, as this space craft will be flying on a ballistic trajectory as far as I know, though it will have a solar sail so I guess it’s possible it could be turned around and used to slow the craft. It will be flying for 20-40 years, but NASA has done that already so it’s not beyond reason they could set it up to automatically be able to do some tweaking to it’s flight once it enters the AC system, though my understanding is the plan is more as a fly by and just send back whatever data it gets before it exists the system.

39

Voyager 1 is not the fastest thing humans have built. The Parker Solar Probe holds that honor. And it is still slower than molasses when it comes to the distances of interstellar space.

40

Also keep in mind how hard it was to find Uranus, Neptune, and Pluto even knowing the rough plane they orbited. We’re still finding Oort objects, IIRC. Perhaps photography and computer analysis will do a better job - but also, if a planet is closer to the probe that it’s sun, it will be mostly the dark side we (don’t) see. However, we’re probably looking for the inner planets anyway so does not matter, except to avoid collisions. While earth orbit may be able to host telescope arrays able to see extrasolar planets, I assume a probe would not carry quite as large an imaging device. However, if we can send a probe to Alpha Centauri, we certainly could build one capable of tracking a planet zipping by. In this case, too close is probably not productive if it results in motion blur…