Remove ads

SpaceX Launches, Landings and News

I think most people in this thread are giving this way too much credit.

waito_x said:
It is not overly complex, just cumbersome. For a majority of these auto-docking procedures to work in space, there needs to be a dedicated virtual conference in place, like a working GPS. This is about relativity offsetting the chronometer and changing the broadcast timings.
Click to expand...
Click to shrink...

This is just word salad. Literally just insane mumbo-jumbo. Either he waito_x is just trolling us, or they have no idea whatsoever what they are talking about.

Relativity has no impact whatsoever on docking in space. The relative speeds between the two objects are measured in fractions of m/s, there will be no "offsetting the chronometer" or "changing the broadcast timings", whatever that is supposed to mean.
 
Tuna-Fish said:
I think most people in this thread are giving this way too much credit.



This is just word salad. Literally just insane mumbo-jumbo. Either he waito_x is just trolling us, or they have no idea whatsoever what they are talking about.

Relativity has no impact whatsoever on docking in space. The relative speeds between the two objects are measured in fractions of m/s, there will be no "offsetting the chronometer" or "changing the broadcast timings", whatever that is supposed to mean.
Click to expand...
Click to shrink...

I think it is more likely to just be a misapplication of knowledge. They know that you need to deal with relativity or you get large scale GPS drift and they are trying to apply that to anything that is moving in space.
 
Tuna-Fish said:
I think most people in this thread are giving this way too much credit.



This is just word salad. Literally just insane mumbo-jumbo. Either he waito_x is just trolling us, or they have no idea whatsoever what they are talking about.

Relativity has no impact whatsoever on docking in space. The relative speeds between the two objects are measured in fractions of m/s, there will be no "offsetting the chronometer" or "changing the broadcast timings", whatever that is supposed to mean.
Click to expand...
Click to shrink...
I think they're talking about remote controlling the thing? Which, yeah, I'm pretty sure isn't done. :/
 
Tuna-Fish said:
This is just word salad. Literally just insane mumbo-jumbo. Either he waito_x is just trolling us, or they have no idea whatsoever what they are talking about.

Relativity has no impact whatsoever on docking in space. The relative speeds between the two objects are measured in fractions of m/s, there will be no "offsetting the chronometer" or "changing the broadcast timings", whatever that is supposed to mean.
Click to expand...
Click to shrink...

I was summarizing without going into much detail. The response immediately after mistook my mention of GPS as some variable to consider, when really I just ask that people revisit how machines track time however infinitesimally using contemporary chronos and given the clock speeds of payload electronics.
 
waito_x said:
I was summarizing without going into much detail. The response immediately after mistook my mention of GPS as some variable to consider, when really I just ask that people revisit how machines track time however infinitesimally using contemporary chronos and given the clock speeds of payload electronics.
Click to expand...
Click to shrink...
Which we shall continue to point out has absolutely no relevance to the task at hand.
 
waito_x said:
I was summarizing without going into much detail. The response immediately after mistook my mention of GPS as some variable to consider, when really I just ask that people revisit how machines track time however infinitesimally using contemporary chronos and given the clock speeds of payload electronics.
Click to expand...
Click to shrink...

That is both a solved issue and also a completely unimportant one as far as docking spacecraft go. One spacecraft's job is just don't rotate and put out a dumb beacon signal. All the maneuvering is done by the other space craft using the beacon just as a point source of radiation. The beacon isn't even required it just makes it easy and simple enough that 1967 era computers could do it. One clock could be a swatch that only read time out in .beats and the other was reading time out in New Earth Time that read time out as degrees with zero synchronization even attempted and it wouldn't matter. As long as ground control puts them within a few km of each other then they can dock.
 
Ciber said:
has absolutely no relevance to the task at hand
Click to expand...
Click to shrink...

Is it this task?
Tasrill said:
All the maneuvering is done by the other space craft using the beacon just as a point source of radiation. The beacon isn't even required it just makes it easy and simple enough that 1967 era computers could do it.
Click to expand...
Click to shrink...

I am honestly concerned that the overall reliability of the autonomous docking maneuvers points to the results of a 1967 duct tape experiment. It does not address the rigors of machine intelligence in the slightest.

We are not dealing with remote-controlled receptor ranges, this is about tasking lightweight / compact circuitry with handling absolute staggering amounts of real-time data. Digitizing the local space environment into a navigable map that an AI 'dummy' pilot can interpret is the challenge here. And at the orbital velocities in question I don't think beacons will do the trick. Please feel free to enlighten me on this.
 
waito_x said:
We are not dealing with remote-controlled receptor ranges, this is about tasking lightweight / compact circuitry with handling absolute staggering amounts of real-time data. Digitizing the local space environment into a navigable map that an AI 'dummy' pilot can interpret is the challenge here. And at the orbital velocities in question I don't think beacons will do the trick. Please feel free to enlighten me on this.
Click to expand...
Click to shrink...
What "staggering amounts of real-time data" are we talking about here? Autonomous docking systems aren't like automotive self-driving systems; even low Earth orbit is pretty empty of hazards and the environment can be heavily simplified because it's the one environment where you can work with "assume a perfectly spherical cow in a vacuum" and get good real-world results. Like everybody else has been saying, automated rendezvous and docking has been a solved problem for decades. It's not the technical challenge of, say, getting a Tesla to drive itself down I-5 during rush hour without causing a 300-car pileup.
 
waito_x said:
Is it this task?


I am honestly concerned that the overall reliability of the autonomous docking maneuvers points to the results of a 1967 duct tape experiment. It does not address the rigors of machine intelligence in the slightest.

We are not dealing with remote-controlled receptor ranges, this is about tasking lightweight / compact circuitry with handling absolute staggering amounts of real-time data. Digitizing the local space environment into a navigable map that an AI 'dummy' pilot can interpret is the challenge here. And at the orbital velocities in question I don't think beacons will do the trick. Please feel free to enlighten me on this.
Click to expand...
Click to shrink...

I have no clue what you are even referencing with some duct tape experiment. The data isn't staggering it is all stuff that was done with on board computers in 1967 as I already linked to and have told you about before. The very first ever docking between a space craft and a space station was done with automatic docking systems. I know of two cases were Russia didn't use it. The first was because the space station was dead so the automated docking didn't work. The other was done as a test to save money and nearly lead to the loss of the Mir space station.

Also what complex map? You are docking and all docking is you and another object with nothing between you doing spherical cow in a vacuum Newtonian math. Even calculating an orbital rendezvous is dumb, basic math with a handful of variables. Space is rather famous for having very little stuff in it to run into or much complex topology.
 
First, to curb any sense of trolling, I waited to post again until several of you had a chance to weigh in. Spherical livestock is a layman's punchline to this thought experiment. Just to stress that the only time the actual science behind self-driving tech and orbital payload metrics approached this level of "anything goes" was when Mr. Musk parked his Roadster in space.

Yes, I get what you are saying, in assuming that once either orbital module or service package is slowing down to several meters per second relative velocities the auto-guidance can manage much better than even a self-driving car and that is saying something. Self-driving mesh network improvements are getting pretty good at dealing with a physical road, but hold on. This is a flattened or otherwise preconfigured corridor with a cross-sectional window of updates with little consideration for up and down. Trying to tether in space is going to delve into the "nth row" of algebraic guess work that I don't see anyone appreciating. The hazard is thinking AI can manage the rest of the maneuver using only a handful of variables from a 1960s experiment.
 
waito_x said:
First, to curb any sense of trolling, I waited to post again until several of you had a chance to weigh in. Spherical livestock is a layman's punchline to this thought experiment. Just to stress that the only time the actual science behind self-driving tech and orbital payload metrics approached this level of "anything goes" was when Mr. Musk parked his Roadster in space.

Yes, I get what you are saying, in assuming that once either orbital module or service package is slowing down to several meters per second relative velocities the auto-guidance can manage much better than even a self-driving car and that is saying something. Self-driving mesh network improvements are getting pretty good at dealing with a physical road, but hold on. This is a flattened or otherwise preconfigured corridor with a cross-sectional window of updates with little consideration for up and down. Trying to tether in space is going to delve into the "nth row" of algebraic guess work that I don't see anyone appreciating. The hazard is thinking AI can manage the rest of the maneuver using only a handful of variables from a 1960s experiment.
Click to expand...
Click to shrink...

A completely empty road is far, far harder than space. If you go off the road then you get stuck or crash. If you get your orbit wrong you just have to spend extra delta V to correct it. Friction is a thing you worry about on earth but is a think that is only a problem in space on the scale of months or years when you are in earth orbit. Roads can have ice, water, or a billion other things that change your traction, response to maneuvers, and on and on. Space only has variations in gravity and drag and both of those are tiny forces when all you are doing is spending 1-5 days trying to match an orbit.

Again what 1960's experiment are you even referring to. Give me a link or something? Literally what duct tape experiment are you taking about? If you meant docking in your original post let me explain in detail. The very first automatic docking was done in 1967. It was then used continuously from then to today with gradual improvements. Every single Russian Soyuz and Soyuz derived space craft since 1967 that has docked has docked automatically except in two cases. Every single time a Russian space craft has gone to the ISS it has docked automatically. Every Russian space craft going to Mir docked automatically except the one time a human actually tried to dock and nearly destroyed the station. Every dock between a Russian Salyut station and a space craft was done with automatic docking except for when the station was dead in space with ice covering every internal surface. You would know this if you had read any of the links you were given. If someone gives you a link as evidence please at least open the link.
 
This discussion is sounding very much like some one ignoring the issue of it already being done, and being done for a long time. I think at this point they'd have to make a proper argument why automatic docking is so hard, despite it being done for half a century already. Because it's hard to take such a point seriously otherwise.
 
At this point I'm 100% convinced that friend waito is conflating self-driving, which is a fairly involved and challenging puzzle for AI researchers in a couple different ways, with automated orbital maneuvering which is... not, it is not that. I'm not entirely sure where they got that idea from, or how to disabuse them of the notion - though their syntax is giving me flashbacks to 1994 sci.space.* so disabusing may not be on the menu.
 
Ciber said:
What if they were recording games to use to train an AI for docking?
Click to expand...
Click to shrink...
This was the original question in the thread that led to some talk of automated docking in space. Where the disconnect occurred is now plain: my original apprehension about the earlier post, that is using recorded flight data to train A.I. for these type of maneuvers. I assure you I looked at the articles, and again emphasize that we aren't there yet. Am I overinflating these points? That is certainly possible. So I want to either give this thread some cooldown or field actual talk i.e. to absorb some counterarguments and build from there.

I know that automated docking with the ISS and earlier stations in orbit has worked repeatedly, and it is good to use those examples especially where space agencies begin to transition from human proficiency to machine learning context.

But once again, using pilot data to help benchmark what is expected of an A.I. instead piloting was the big no-no. That is where I was basing my concerns. So I want to ask, are we describing these 1967 and later flight control mechanisms (and yes, these have then had incremental improvements, good point a moment ago @Tasrill ) as the gold standard for transitioning to a robust machine learning pilot replacement framework?

edit: I will revisit the other point from Mal-3 following someone else's chance to reign this discussion in to something approaching a discernable train of thought. Bless.
 
Last edited:
waito_x said:
This was the original question in the thread that led to some talk of automated docking in space. Where the disconnect occurred is now plain: my original apprehension about the earlier post, that is using recorded flight data to train A.I. for these type of maneuvers. I assure you I looked at the articles, and again emphasize that we aren't there yet. Am I overinflating these points? That is certainly possible. So I want to either give this thread some cooldown or field actual talk i.e. to absorb some counterarguments and build from there.

I know that automated docking with the ISS and earlier stations in orbit has worked repeatedly, and it is good to use those examples especially where space agencies begin to transition from human proficiency to machine learning context.

But once again, using pilot data to help benchmark what is expected of an A.I. instead piloting was the big no-no. That is where I was basing my concerns. So I want to ask, are we describing these 1967 and later flight control mechanisms (and yes, these have then had incremental improvements, good point a moment ago @Tasrill ) as the gold standard for transitioning to a robust machine learning pilot replacement framework?

edit: I will revisit the other point from Mal-3 following someone else's chance to reign this discussion in to something approaching a discernable train of thought. Bless.
Click to expand...
Click to shrink...


You never talked about machine learning or anything like that but keep talking about clock cycle and time keeping differences or it being massively complicated. Those things aren't really related and are massively different things to critique. The idea of using machine learning for a docking system is dumb because machine learning outputs unreadable code that is a black box while docking and even astronavigation are simple problems that humans can handle. Machine learning is used on cars because computer vision, sensor fusion, control on a highly variable surface, all while having a fraction of a second response time is far beyond what a human can code and test. In space you are either very slowly docking with plenty of time to react, doing orbit change burns if something big enough goes wrong you just stop doing the burn and spend the rest of the orbit having ground control see what went wrong. Cosmonaut Dzhanibekov expert piloting to dock with an unpowered station would be extremely easy by modern ai standards honestly unlike the challenge of telling if something is a plastic bag or a solid object in a fraction of a second.

Also in the context of 1967 automatic docking was AI technically :V
 
Tasrill said:
Cosmonaut Dzhanibekov expert piloting to dock with an unpowered station would be extremely easy by modern ai standards honestly unlike the challenge of telling if something is a plastic bag or a solid object in a fraction of a second.
Click to expand...
Click to shrink...

Can you elaborate on the plastic bag - is that regarding detritus on a highway for self-driving A.I.? Anyway, thank you for the clarification on the other points.
 
waito_x said:
Can you elaborate on the plastic bag - is that regarding detritus on a highway for self-driving A.I.? Anyway, thank you for the clarification on the other points.
Click to expand...
Click to shrink...

Being able to tell a empty plastic bag, pile of leaves, or so forth from a solid and potentially dangerous object is very important for a self driving car. Sudden changes in direction or slamming on brakes is dangerous as it increases the chance of losing control or being hit by other traffic so if your self driving car treats anything on the road the same it is going to be so cautious that it is actually dangerous. It is also a hard problem as humans get that kind of thing wrong to. In space a massive world spanning radar system either tells you that you are going to hit something hours in advance or you hit it as there is no way to actually move out of the way of something by the time you can see it.
 
Enough'a that nonsense! Today's the day, sports fans!

Well... it might be the day. The biggest constraint right now is weather; the weather at the launch site has improved to 50% go conditions, but weather along the ground track has deteriorated. Specifically, a good sized chunk of the track in the launch abort zone now has a tropical storm sitting right off it. NASA are currently meeting to determine if they're gonna call it off for today or not.
 
Bit late...but as far as docking and collision avoidance are concerned, spacecraft don't need to talk to one another explicitly. Sensor data and the right control protocol can be used to have them basically avoid one another without any communication needed, therefore making latency moot as long as they're autonomous (and not controlled from some Earthside base station). The source paper for this is van den Berg's AVO which cites older NASA work on purely velocity-based avoidance.
 
I really hope this goes well. With any kind of first crewed launch like this I am worried about things going wrong.
 
Good news: The crew is aboard Dragon and everything's buttoned up, no problems with spacecraft or launch vehicle.

Bad news: It's currently raining like twenty motherfuckers and some moron who thinks he's Tom Cruise just buzzed the launch pad. Weather is very much the big "if" here and right now it isn't looking great. We'll probably get the final call at around T-45:00 before propellant loading begins.

Next window opens on Saturday at around 3:30ish EDT IIRC.
 
And if they scrub Saturday too then the backup for that is Sunday. After that it might be a week or so. Looking into it, after Sunday it's probably Tuesday.
 
Last edited:
You must login or register to reply here.
Remove ads
Back
Top