I find it truly amazing that New Horizons was able to take photos of a 20 mile object whilst flying past at ~8 miles/sec - that’s an incredibly small space and time window within which to capture the images. Does anyone know how NASA manages to pilot spacecraft with this level of precision? What kind of engineering processes do you need to enable this?
Actually doing the fly-by is less impressive (the fly-by was planned and programmed a while ago) than the fact that we were able to line up Horizons to an object like this at all!
I find it more impressive and awe inspiring that stuff in space is so predictable that we can actually program the fly by such a long time in advance and get such a high degree of accuracy.
I also find this somewhere between impressive and amazing. However...
...after watching the Dragon/ISS docking manoeuvre for quite a bit longer than I probably should have, I think I now understand this precision better. When we think of "flying" including flying in space, we automatically have an image of something that's a bit unstable, always being buffeted a little, because that's how all the flying we are familiar with is.
Actually seeing that docking manoeuvre, you get a feel for just how stable things are in space. When the Dragon is parked in its relative position (10m?) it is just there, rock solid, as if attached with steel beams. No, better, as if both parts are part of a single piece of granite. No quivering, nothing.
Of course this is intellectually clear/obvious, but seeing it in practice is something different.
Yup. And if you've ever done maneuvers manually in KSP, you know that a tiny error in your nudge early on - e.g. hamfisting the throttle, or being a little too late in mid-maneuver staging - leads to a huge difference farther out, which you then have to correct, wasting precious ∆v :).
I'm very much impressed by the control precision of real-life space probes.
Wow, I actually had no idea; how is this all pre-planned, does somebody at NASA just put "67-P/Churyumov" into a "Orbital Maneuver Planner" and let some computers crunch the possible paths out? Or does someone actually sit down and come up with the possible sequence of orbital assits (to later verify with a computer) with a pencil and paper?
It's a combination of both. They have tools that help worth the trajectory calculations, but they have to manually decide on the basic layout of the path to take.
There is a very interesting interview with Pablo Munoz from the Bepi Colombo team about flight dynamics on the Omega Tau podcast that explains this: http://omegataupodcast.net/295-bepicolombo/
The other interviews on the same episode are also worth listening to. In fact, the entire podcast is great.
Common orbits are solutions to the two body problem, but Shane Ross has a few videos on youtube about chaotic solutions to the three body problem, allowing objects to move in very complex orbits with very little fuel. This has been used for a few missions as I understand it, the math is WAY above my head, but the video is still quite watchable with lots of cool orbit animations:
I think everyone with an interest in spaceflight should play Kerbal Space Program. It gives you a sense of scale, both of space and time (luckily you can fast-forward to 100.000x), and also teaches you the basics of making orbital trajectories and correcting them, landing on planets/moons with different gravity... I'm really glad to have played it!
> photos of a 20 mile object whilst flying past at ~8 miles/sec
Not to detract at all from the astonishing technical achievement, but that's the wrong comparison. You can see the ISS with your naked eye, and it's a <1 mile object flying past at 4.8 miles/sec. The right comparison is with the distance at which this photo was taken: 18,000 miles, or about half an hour from closest approach.
Actually, the right comparison is of the closest approach distance (~2000 miles) with the distance from earth (4,000,000,000 miles). That's like launching a missile from Los Angeles to New York and hitting your target to within a meter.
To be fair, while that's an accurate representation of the distances, it's not accurate of the difficulty. They don't have to deal with unpredictable atmosphere. There's reasons laser-guided missiles are about 1-2 meters-ish of accuracy with constant guidance, while we can basically fire-and-forget spacecraft to 4 billion miles away.
No, that's actually not true. Hitting Ultima-Thule is much harder. For starters, just figuring out where U-T actually is is very difficult. The observations we have only constrain two of its three degrees of orbital freedom. The third has to be inferred via orbital mechanics. Likewise, figuring out where the spacecraft is is also non-trivial. By way of contrast, figuring out where you are relative to New York to within a meter of accuracy is, nowadays, a simple matter of switching on your GPS receiver. (Even without GPS, all you need is inertial guidance that is stable for half an hour or so. That's much easier to achieve.)
Finally, there are unpredictable orbital disturbances. N-H uses thrusters for attitude control, whose effect on the trajectory is not entirely predictable. (c.f. https://en.wikipedia.org/wiki/Pioneer_anomaly, which was only detectable because, although Pioneer had thrusters, they were turned off for long periods of time).
I know these things because I used to work at JPL. There are entire teams dedicated to spacecraft navigation. The stuff they do will blow your mind.
I wonder how much would it cost to build a GPS-like network for the whole solar-system...
The good thing is that, for such a thing to be really useful, we'd have to invent better propulsion systems which would, in turn, make the network much cheaper to build.
Pulsars can be used to arrive at celestial positioning accurate to ±5 km. There is a test rig on the ISS that has validated this. No need to build human made beacons when the universe has so graciously provided us with them.
They plan to get within 1km, which is quite impressive for one system alone. Thanks for the pointer (after reading it, I remember having read about NICER, but didn't remember the precision they got)
Id guess it could be done for not that much, assuming accuracy can go from a few meters to few thousands of miles. (Someone more versed might be able to give a better oom on accuracy). Just need to launch 4(+) satalites, 2 up and 2 down relative to the orbital plane. So long as a ship isn't hiding behind a planet/sun it should have LOS to all 4 for triangulation.
Not exactly. You can use star location to determine a spacecraft's orientation (and they do), but its location is determined by measuring the doppler shift of the carrier signal sent from the spacecraft back to earth and running those results through some pretty hairy math. GPS works by having a network of beacons in known locations transmitting signals with known timing. I just learned from a sibling comment in this thread that they're working on using pulsars as the beacons. But building an artificial GPS system for the solar system would require sending out a network of satellites into solar orbit. That might be technically feasible, but not economically feasible.
Actually, in the case of New Horizons, it is. Mission navigators use LORRI images of known objects relative to the background stars to calibrate their position.
Brain wedgie. I meant six. And direct observations give you four of the six. I guess I was just thinking about 3-d location and just forgot about the time derivatives. Sorry.
"They don't have to deal with unpredictable atmosphere"
Well, no, but they do have to deal with an atmosphere that has, until they get to it, only been theorized about, never observed - which is not the case for the atmosphere above New York.
There have been precisely four spacecraft that have gone anywhere near so far out as New Horizons, and they are between them the source for the majority of data we have about the atmosphere in the outer solar system. One of them discovered an anomalous acceleration effect that confused scientists for decades (and which, while it turns out not to have been externally caused, certainly left scientists wondering for a while whether their model for the solar wind was accurate). Of those that we believe have passed through the termination shock of the solar atmosphere, none have yet returned much accurate information, so we actually don't know much about the conditions there. I don't mean to say that these effects have meaningfully impacted navigation for deep space probes, but more these are genuine voyages of discovery - where they're going is predictable but until they get there we really don't necessarily know whether our predictions will be accurate.
> The right comparison is with the distance at which this photo was taken: 18,000 miles, or about half an hour from closest approach.
Implicit in this statement: the camera was pointing close to the direction of travel and the target wasn't moving much within the field of view. Hugely easier than trying to image from the side at closest approach, which at best would have given a smeared image and at worst a complete miss.
It's still impressive and awesome. But always try to skew the odds of success in your favor when dealing with stuff like this. You have one pass and then the opportunity is gone.
the development of double precision floating point, and the kalman algorithm, were two major improvements. Basically there is a computer at JPL that, given a location in the solar system and a current location of a ship, will give you instructions on when to burn. After you burn fuel to move, you update based on visual locations of known stars, fit a model, and do incremental burns until convergence.
The book Digital Apollo touches on a wide range of issues associated with building systems that can do this.
New Horizons is a mere few billion miles/km away, while financial matters these days sometimes involve trillions. Plus extra digits for cents and fractions thereof, of course.
“There are 10^11 stars in the galaxy. That used to be a huge number. But it’s only a hundred billion. It’s less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers.” - Richard Feynman
i know you're joking but it's really the precision at large numbers that makes doubles so valuable. Typically when somebody is steering a spacecraft to a remote location, the large distance as well as the fine precision needs to be represented. I am not aware of people working with very large sums of money who also need penny-level precision (and even then, floats aren't the right solution).
JPL also has PhD applied mathematicians on staff to ensure numerical stability of their algorithms at double precision. Most developers do not have the ability to do that sort of non-trivial analysis. They are not just blindly saying "a double is good enough."
The problem with floating point for money is that simple calculations can give different answers than you would like because of how numbers are represented. E.g in a JS console:
> 0.1 * 0.1
<- 0.010000000000000002
This matters whenever you do things where the error might be allowed to accumulate and you're not careful to control for it. The general advice to avoid floating point for money is not because it's impossible to do correctly, but because it's very easy to get wrong in ways that are hard to discover with testing, and doing it with money is one of those areas where it's easy to get it wrong in ways that people will care about (because you're suddenly paying the wrong amount of tax, for example).
I don't know about you, but I don't trust myself to get this right, much less developers who often don't understand the issues involved.
It’s good enough for dollars and cents if you put thought into it, but you can’t just represent money as double-precision dollars and expect to get the results you need.
floating point was designed for scientific and numerical/math applications (written by people with numerical analysis skills), not money. There are decimal data types which are much better for money management.
I searched for "New Horizons GNC" and this link came up [1]. "The AOCS/GNC (Attitude and Orbital Control System / Guidance Navigation and Control) subsystem takes care that a spacecraft points to a specific point in space, determines the s/c attitude, and does trajectory corrections."
They actually managed to predict the bi-lobed shape by keeping track of when the shadow of the object swept different points on the Earth’s surface. Amazing.
I watched that video and when the pictures finally came out i was completely floored by how accurately they were able to predict the shape. An incredible problem solving strategy that seems very simple in hindsight.
Re: able to take photos of a 20 mile object whilst flying past at ~8 miles/sec...Does anyone know how NASA manages to pilot spacecraft with this level of precision?
I cannot speak for this probe, but other probes have made the probe itself or camera boom move slightly to compensate for the target moving relative to the probe. It's kind of like moving your head back and forth to follow a tennis match if your eyes alone have difficulty tracking.
The object looks like BB-8. We'll get even better pics in the coming weeks.
In the press conference they said that the spacecraft tried to take one megapixel image of Ultima Thule one hour after this image. Interesting to see it in couple months.
