Monday, July 13, 2009

Subway Accidents and Space Pilots

Three terrestrial transportation mishaps recently in the news have called attention to the challenges facing human operators of highly automated systems. In Washington DC a subway train under automated control collided with another train, killing several people including the key witness, the train operator aboard the moving train. [Link may require free registration.] At about the same time, the investigation of a 'commuter' airliner crash in upstate New York concluded that the pilot failed to respond to a critical control input during a semi-automated landing approach in bad weather. And a highly automated Air France Airbus crashed in the mid-Atlantic after a cascading series of computer failures.

The first and last of these accidents are still under investigation, and unless the 'black boxes' are located and recovered from the ocean floor we may never have a full picture of what happened to the Airbus. The commuter plane crash was due to pilot error, and the training and working conditions of the crew may be implicated. But all three accidents seem to involve the human-machine interface – specifically, what happens when humans who normally are only along for the ride suddenly have to take charge.

A previous Washington subway collision investigation revealed that train operators had been specifically directed not to take the controls themselves. Whatever happened this time, we can guess at minimum that the operator had not been controlling the train during its pre-emergency run, and either did not recognize the developing emergency in time or had been told to ignore potential warning signs (such as trackside signals).

The commuter plane's control column began vibrating, a stall warning, and the pilot 'instinctively' pulled it back – putting his plane into a full stall – instead of pushing it forward to get the nose down and regain airspeed. The Airbus was at cruising altitude, the crew almost surely not handflying it, when computers went out. We can only guess that the cockpit became a desperately 'busy' place, the crew frantically trying to fly the plane on backup instruments in turbulent weather at night, while also trying to reboot the computers. In all these cases, the humans had (probably) been largely hands-off … till all hell broke loose. The resulting problems have been dubbed the automation paradox.

I have experienced a minor form of this myself, while driving on cruise control on a largely empty highway. A sharper curve comes along, or enough nearby cars appear to become 'traffic,' and I suddenly have to go off cruise to change speed. The experience is slightly jarring – and in these cases I had my hands on the wheel, alert, already involved in driving. It would be far more jarring if the car were fully automated and I wasn't driving at all until I suddenly had to take control.

The implications of all this for space operations are fairly clear. Most of the time, in space, there is no call for 'piloting' of any sort. Spacecraft on orbit stay on that orbit. Even when accelerating, especially with a milligee deep space drive, there is rarely much reason to take the controls. But rendezvous and docking are a different matter. I suspect that as space travel becomes common, a common type of accident will be low speed collisions – which can easily be catastrophic, since 'low speed' is up to tens of meters per second, comparable to highway (or subway) speeds.

One such low speed collision has already been experienced, happily without catastrophic results, when an unmanned Progress supply craft banged into the Mir space station. In that case, as it happens, a cosmonaut aboard Mir was handflying the Progress by remote, and misjudged the docking approach. But avoiding all handflying is no perfect solution. I have done enough coding to know how elusive bugs can be, and there is also such a thing as hardware failures.

So how do we deal with the automation paradox in handling spacecraft in close proximity to each others, or other situations (such as planetary landings!) where sudden emergency control inputs may be called for? It goes without saying that the humans involved need to train and practice their emergency procedures. But I am inclined to suspect that that positive handflying may be the better procedure, so that if an emergency arises the human is already in the loop, and doesn't suddenly have to shift their whole mental set to engage in emergency response.

This does not preclude automatic override, e.g., a retroburn if a docking approach goes out of parameters. This in turn can be over-overriden by the pilot, if the override system itself goes haywire. The control station will be a busy place, but with humans who were already actively engaged in the process.

The comment thread is open for discussion!

Related link: Comments on my recent piece about the Singularity led to a discussion of automation, AI, and their implications. And by the way, if you're not reading the comments you are missing a lot of the substance of this blog!


Kedamono said...

This sort of follows what I've predicted for automated trucks. I think that the first fully automated civilian vehicles will be long haul trucks. For one thing they wouldn't suffer from driver's fatigue over the long distances, and will obey the laws of the road, reducing any fines.

But... There will be a "driver" on board, if only to do maintenance, fill the fuel, tank, unload cargo, and in those rare situations, guide the truck into a difficult to get to loading dock. (I know of one right down the street, where the truck has to backup across a lane, over and down a hill and navigate a narrow alleyway. It's hard for human drivers, I think it might be impossible for automated systems.)

The driver is aboard because the Teamster's Union will demand it, as would the insurance companies. Figure that there will a few attempts to sabotage some of these trucks "to protect union jobs", but they will fail as these trucks will have "black boxes" and show that the human was at the wheel.

How does this figure into Rocketpunk?

Robot freighters are de rigor in SciFi, plying the spacelanes from one world to another, with nary a human on board. How hard would it be to rewire or program one to become a WMD?

The only way to make sure that doesn't happen is to have a crew on board. Their pay isn't that high, since for the most part they are not working. Only during their shift in the control room would they be considered to be working.

And unless the shipping company ran drills on a regular basis, the problems that afflicted the Air France pilots, would happen to the crew of the freighter.

And the consequences of a fully laden spacefreighter impacting a world would be far greater than a plane crashing.

Citizen Joe said...

There is an additional problem in landing whereby the shift from zero gee to full gravity causes you to over reach or under reach until you get used to the new gravity condition. During landing is not a good time to acclimate.

I think that when it comes to docking with space stations, the station would send out its own pilot to rendezvous with incoming craft. After an inspection of the ship's capabilities, pilot would then guide the ship in using the station's parameters.

That being said, I've also designed huge mass drivers for launching and catching ship in orbit around Mars and the Moon.

Calsir or Carlo, whatever said...

2 points. The operator on an automated truck is much more useful for liability than for emergency control, since the company can blame him if something goes wrong.

Secondly, docking accident will not happen at "tens of metres per second" (which is more or less the speed of a car accident) but rather on the order of "tens of centimetres per second". If relative velocity is above those values, you are not trying to dock, but to ram... It is more or less like stating that a 90° nose dive from cruise altitude is a "landing accident" :). However, your point still stands.

We must not forget that that much automation also reduces the risk of pilot error: this is the Airbus doctrine that has the aircraft to protect itself from damaging input (no, I don't work for them).

Of course, in the unlikely event that the 4~5 redundant control computers fail, the vessel is quite fubarred, but it is still way less likely than a pilot error.

Rick said...

Kedamono - Quite plausible that long haul trucks will be the first automated vehicles. But in the scheme as outlined they could be very much subject to the same problem as the DC subway.

My approach (tentatively) would be something like this: Define a safety box around the truck. The driver drives, keeping the truck centered in the box. If it strays outside parameters, first an alarm goes off, then the automated system takes control - with an override available if the system is screwing up.

But I am inclined to keep robotic freighters, because putting a crew aboard isn't just a matter of hiring a member of the Interplanetary Union of Spacers - it means installing a costly and massive life support system. (Especially if cargo craft take economical but slow Hohmann orbits.)

Orbits will be planned so that simple drive failures won't put them on a collision course with a planet (or hab, etc.). If major failure - or sabotage - deflects it onto a dangerous trajectory, there should be plenty of warning time to scrag it.

Citizen Joe - Good point about landing. I'm not sure whether this is an issue, but we won't really know till we are doing it.

I don't see a need to physically send pilots out to board craft; they can handfly it in by remote. If the control linkage itself goes down you have a problem, but that is true for an onboard pilot as well.

Calsir - Nice point about trucking companies pinning it on the driver! Though eventually the law should catch up to this situation.

Yeah, you're right about 'tens' of m/s, though the initial stage of a rendezvous approach could be at around 10 m/s. (In the Mir incident, the Progress made initial approach at 6.5 m/s instead of 5 m/s as intended.)

I don't disagree that automation generally reduces pilot error. LA Metrolink had a much worse rail accident last year - 25 dead - because it lacks automated emergency overrides, and a train operator ran a signal. (Though there's been question raised about the red signal aspect's visibility.)

As I understand it - admittedly from the mass media, none too reliable on technical subjects - Airbus and Boeing have significantly different design philosophies in their latest generation aircraft, with Boeing calling for significantly more flying by the crew.

This might not have made any difference for the Air France crash - planes haven't normally been handflown during cruise flight at altitude for many years. (Not to mention that we have minimal idea of how the emergency developed, and none at all of how the aircrew responded.)

Citizen Joe said...

Automated vessels and 'thrown' cargoes would need to have transponders and registered flight paths. If the transponder goes down or it diverts from its registered course, then it would be considered abandoned and thus salvageable. If the company can regain control of their vessel/cargo, they could file another registration (so long as the new course is not a navigational hazard). That is where the space patrol comes in. There may be the rare instance where they might save a ship, but mostly they need to clear any wayward objects. Assuming the object is not actively maneuvering, it is a simple matter to plot a targeting solution out to many light minutes for a certain 5 mile long X-Ray laser ship. Of course, blowing it up is not the objective, but rather targeting a specific point which would vaporize into a jet and modify the course slightly.

The gravity change issue is something that the Mars Mission astronauts are considering right now.

Rodney said...

I think there’s another issue to consider: trust. How often have you been on a long road trip, riding along at about 70 mph (or faster depending on how brave you are), and look at the guy next to you looking as if he’s about to fall asleep at the wheel? He might have cruise control on and he might be on a very straight road, but you still would like him off the road as soon as possible. I have, on several occasions, backed away from someone looking as if they were a few minutes from an accident.

The point is that, if the first civilian vehicles fully automated are long haul trucks, I don’t think the general public will trust them. They might know that the computer is in control, and they might believe that the computer is very reliable, but the average driver would like to know that the driver is actually, well, driving. I like Rick’s idea about the “safety box,” and I think that’s more likely than full automation.

My thinking is that these attitudes will probably carry on for a while. If it ever gets to the point where the average person deals with a computer on a daily basis that has a normal functioning state of “excellent” rather than “failing as designed,” then the acceptance of robo-freighters will be a given. I have always had a problem with the idea of removing the human from a spacecraft that’s designed to do more than take pictures of planets, however, and I think the general public wouldn’t feel comfortable with something carrying a great deal of kinetic destruction speeding around their planet.

Unfortunately, that’s what would be the biggest obstacle to robo-freighters. The general public, guided by apathetic feelings toward research and listening to the media looking for a good story (and good news doesn’t make a good story) might force whoever represents them in a law making capacity to make sure that robo-freighters never happen. Even if, say, Federated Boeing proved that their Cargomaster 3000’s computer never failed(unlikely I know) and had bug free programming (again unlikely), if the public heard about the story where the Cargomaster X test model crashed into the moon, there would be no robo-freighters.

Of course, this depends on the understanding of technology and the public’s attitudes toward such things, so this could change in the future.

Rick said...

Citizen Joe - Actual salvage is fairly unlikely unless the failure occurred near the start or end of a mission. Similarly, most robotic spacecraft that go awry can simply be cataloged, much as we already do with space junk. But on rare occasions salvage will be feasible, and on similarly rare occasions action will have to be taken to swat something aside.

Rodney - good point about the trust issue, certainly as applies to road traffic.

Regarding robotic spacecraft, the public is already accustomed to them! So apart from passenger craft (and barring a spectacular disaster), I don't think there will be the problem with robotic craft in space that there would be with robotic trucks, or large drone aircraft. It is 'grandfathered,' so to speak.

For comparison, if subways and rapid transit were new inventions, would we allow crowded high platforms with no barriers to keep people from falling in front of trains? But it's been the norm for more than a century, so people don't think twice about it.

Calsir said...

Yeah, that's the name I use here :).

2 more points:

When we talk about "control computer" we should talk about "control computerS", with different architecture and OSes. It's all about redundancy.

@Rodney, computers for automation perform much better than general purpose computer because they are specialized with a limited set of instructions. Of course, navigating through traffic is much more complicated than keeping a pilot from breaching the flight envelope of an aircraft.

@Citizen Joe: If the radio transponder is critical to mission/vessel safety, you install more than one.

Related: to avoid dangerous collisions with earth orbiting vehicles, cargo vessels could be required to set up a coasting course which has a perigee above GEO: in case of failure, they would simply swing by in hyperbolic orbit. Course correction could be made just before entrance in the SOI.

Have you lads tried orbitersim?

Citizen Joe said...

There are several ways to handle the transfer situations.

1) Ships land on the planet, unload/load, then take off again.
2) Ships enter orbit, then shuttles ferry cargo back and forth.
3) Ships dock at orbiting space station, unload/load, then depart from station.
4) Ships enter parking orbit, then shuttles ferry things back and forth to station.

When you introduce the shuttle, you create more chances for things to go wrong, but the shuttle pilots are very well practiced, thus reducing the chance of a problem, they also minimize the risk. A shuttle bumping into your ship isn't going to be as bad as a huge ship bumping into a space station at comparable speeds.

Landing on planets, either the whole ship or by shuttle, eliminates the danger to the station and catastrophic impacts in space, but it causes significantly more strain on the ships and requires some sort of heavy lifter to get off the planet.

The overall safest method, I suspect would be number 4. It is also the most complex and likely to have the most incidents, but all minor. Fender benders instead of Hindenburgs.

Anonymous said...

One thing I've learned about complex mechanical/electronic/computer systems is that no matter how seldom they experiance a major fail, when they do they do so at the most inconveninant time possible... because unless it is a catestrophic failure, even a minor failure at a critical time is a big deal...and can be the most exciting moments of your life (and NOT in a good way); so, I believe that people will be onboard transport spacecraft (pilots and flight engineers both), to have positive control during critical periods and to have the benefit of a human's intuitive approch to troubleshooting software and hardware problems, as well as humans' creative solutions to problems that automated control systems are incapable of. Whether they are drivers on the interstate or the harried controlers at an out-of-the-way space station, most automated systems have trouble dealing with the nonlogical actions of humans that other humans would take in stride...probably with a lot of cursing, but handled without suffering a 'logic lockup'...I'm sure that fully automated (i.e. robot) freighters will have a place, but only, at least at first, for lower priority cargo and on lower risk trajectories. I've learned, over a 20 year period of maintaining complex communications systems, that nothing man builds is perfect, and that humans 'in-the-loop' is the ultimate backup. I don't see that changing anytime in the foreseeable future.

As far as transfering passengers and cargo from interplanetary transports, the exact method might depend more on what is being transfered and the circumstanses of where the transfer is taking place, than any standarized one-size-fits-all proccedure.

Kedamono said...

At some point, driving a car/flying an aircraft/piloting a space ship will become semi-automated through "Fly by Wire" systems. Air Bus aircraft and mostly fly by wire, so there is no direct control of the craft by the human pilot.

Spaceships have always been fly by wire, but with redundant wires, so that a pilot can do a deadstick landing or control the ship if the computer troika goes down. I'm not sure if Air Bus aircraft have this ability for manual control. As long as you have power. No power...

What it boils down to is that you are not flying/driving/piloting that vehicle, you're telling the computer through a series of inputs how to fly/drive/pilot the vehicle. Any form of catastrophic failure, and you're in a uncontrolled ballistic missile and it doesn't matter if there is a pilot at the wheel, he can't do jack.

Rick said...

These replies are all related, so I'll deal with the issues raised as a group.

Even before fly by wire, aircraft (and other vehicles) depended on mechanical linkages. If the control input fails, the pilot is controlling nothing. In space there's more chance to make repair than in aircraft - unless there's no time, or the damage is non repairable. As Kelsey on Firefly would say, sometimes things are just broke.

Transfer procedures will no doubt vary. If cargo is in pods, the pods can be transferred with no direct docking of vehicles. For passengers you probably want a direct connection - going from liner to pod to station adds failure opportunities for things like airlock.

The choice of manned or unmanned cargo transport largely comes down to cost. In current practice, for example, we lose about 5 percent of unmanned missions, and the 'insurance surcharge' is cheaper than the additional cost of a) providing life support and b) upgrading the whole thing to a safe enough level for human transport.

I suspect that for most cargo transport this will continue to be the case for a long time to come, perhaps indefinitely. It is not unlike the argument for robotic warcraft - you'll lose more, but they are much cheaper to lose.

Anonymous said...

Major bridge disasters tend to occur in cycles. I can't remember the exact frequency and I'm feeling too damned lazy to google it myself, but it comes down to a simple problem. The first engineer or team to design a certain form of bridge does their absolute best to build a stable and solid bridge, but doesn't really know all the stresses that said bridge will face. And after a few years said bridge - And all the similar designs put in place at the same time - will start to fail under the cumulative effect of unanticipated stresses. Repair crews and engineers will fix the old generation of bridges and design a new generation to better withstand the stresses that caused the problems. But they will fail to anticipate a different set of stresses...

Something similar goes on with airplane and train design, and probably ships and buildings as well.

A lot of my friends and most of my coworkers are various forms of engineer. And of course they studied systems failure in school, and those classes all went into detail as to the nature of the problem: Unanticipated stresses, unplanned problems, 'unknown unknowns' in the wisdom of Cheney. And in those classes, my engineering friends tell me, there was almost always one clever person who insisted that the way to deal with the problem of unknown unknowns was to simply anticipate the problem and design for it.

I'm told that these were not stupid people. In a lot of cases these were apparently the smartest people in the class. The fact that they were smart people was part of the problem, because they didn't see how a sufficiently smart person could possibly fail to account for mere mechanical problems. The fact that Leonardo da Vinci, World's Smartest Italian, designed a bridge that failed to plan for the harmonics of people walking over it simply did not register for these people.

So I'm picturing a 50-ton cargo shuttle plowing through a dock at 15 mps, driving its way through the cargo/sorting area and into the manufacturing plant, and finally firing its braking thrusters 45 seconds too late while resting in a vat of chlorine triflouride...

And then a spokesperson for the shuttle manufacturer goes online and explains that the crash obviously couldn't have happened the way it apparently happened, because the company's advanced software suite had contingency plans for all situations.


Rick said...

I think the 'unknown unknowns' line was Rumsfeld's. I have a low opinion of him in general, but I can't deny that that particular line was sound.

And yeah, good classical hubris will always assert that we can plan for everything. We can't. The most we can - maybe - do is realize that we can't anticipate everything, and try to provide a margin of robustness that might give you something to work with when the unknown unknowns come along. Like allowing 50 percent more time and money for a project than it 'should' require, as a cushion.

All of this seems somewhat related to something I read about some years ago, 'normal accidents.' The premise was that as systems get more complex, trying to failure-proof them just supplies more points of failure. (Such as 'false positives' from warning lights, etc.)

Kedamono said...

I think it comes back as to how these robot freighters are programed or not.

My own preference is for the AI/expert system to be taught how to the fly the freighter, basically taught the skill. This skill, depending on how the AI/expert system's hardware is configured, may or may not be transferrable to another computer. It's basically an Apple vs. PC issue. Apple AIs are more user friendly, and load on the artificial personality, while PC AIs are faceless entities, but very good at specific jobs, but are loaded with legacyware.

They can trade data, but not the actual machine code/thought process since it runs on a different CPU.

Anyway, each freighter would contain a clone of the original AI/expert system and use that to control the ship. The learning process is still on, so as time goes by, each ship develops a unique personality and skills from it's trips.

I wouldn't waste an AI on a tedious cargo run, an Expert System would suffice, but for starships, you'd want a full blown AI that has passed all it's piloting tests and is certified for flight. And it may be a fairly dedicated AI to ship's ops and flight, so's not to let it get caught up in questions like lying to the crew about the ship's ultimate mission and killing them while they were in cryosleep.

None of that. Nope. The AI would be like an idiot savant when it comes to it's skill set and personality.

Jim Baerg said...

Citizen Joe said...
"There are several ways to handle the transfer situations.

1) Ships land on the planet, unload/load, then take off again.
2) Ships enter orbit, then shuttles ferry cargo back and forth.
3) Ships dock at orbiting space station, unload/load, then depart from station.
4) Ships enter parking orbit, then shuttles ferry things back and forth to station."

Economics will rule out #1. As Heinlein pointed out in one of his early stories, the wings or heat sheilds & high thrust engines needed for a surface to orbit & back shuttle will be so much dead weight for the interplanetary transfer.

I think that docking ports on space stations should be on fairly long booms or tethers so that a botched docking won't damage the main mass of the station.

So I see the transfer station in low equatorial orbit having tethers a few 100 m to a few km long extending toward & away from the planet. Shuttles from the planet dock at the low end & ships from space dock at the high end. Some sort of machinery pulls the spacecraft toward the center of the station after docking for cargo & passenger transfer. Ships leaving the station are released from the ends of the tethers.

Rick said...

Kedamono - Basic interplanetary freighters shouldn't need anything like an AI. After all, we've flown those missions already, the first generation with 60s vintage computers. (!) And most of the real work will be done at each end, handled by remote from Mission Control. Probably aboard the stations, with whatever mix of AIs and human controllers.

Starships, and complex missions in general, are a whole 'nother matter, and the place where I'd expect to see sophisticated AIs in play.

Jim - Pretty much agree, though perhaps instead of a tether a boom structure. Rather like a present day launch tower, but much more lightly built! For human ops there'd be an access tunnel and 'green room.' In a really bad hard docking you lose the boom and its facilities, but not the whole damn station.

Citizen Joe said...

Don't forget about the fact that you don't pull the ship to the station you pull both together. Large ships could literally pull the station out of orbit with those tactics.

The tether/grapple system would actually work fairly well at L1 or L2 Lagrange points. But you want to come in on a tangent to the orbit, not above or below it. A little extra force forward or back would get absorbed by the Lagrange point and bring the station back to its stable orbit. You could then drop away by thrusting off the tangent and letting gravity pull you clear.

I believe I designed a couple huge mass drivers at the sun-mars L1 and L2 points. Incoming packages would be drawn into these huge rings to slow them down for mars orbit or launch. The rings would then use the down time to oscillate back into stable orbit at the L-points.

Rick said...

The momentum won't be 'absorbed;' the mass driver would oscillate indefinitely around the Lagrange point, or until some other force was applied.

If you have regular interplanetary trade by slinging the baggage, clever planning might be able to balance things out. It would give a whole new meaning to 'balance of trade.' :-)

Calsir said...

@Citizen Joe:
If a ship is in stable orbit close to the station, their centre of mass will be in a stable orbit as well. If you use tethers, the orbit won't change at all. Some fellow countrymen of mine tried the tether concept on the space shuttle some time ago, with mixed success.

Another thought on the transfer to orbit issue: if the cargo has to go on the planet, you want the transfer to happen in LEO, due to the extra cost of going higher. If the transfer is to happen in LEO, you don't want huge space stations, since it would require a lot of fuel for stationkeeping, due to atmospheric drag, especially if you have outstanding booms, solar arrays or radiators. Therefore, it is my opinion that SSTO space/aircraft should be used to directly service transport vessels. Of course, I don't deny that the space station has many logistical advantages that might outweigh its operating costs.

@Rick & Citizen Joe.
If you send anything to an unstable Lagrange point (L1, 2 and 3) with the right trajectory/energy, it will eventually be captured by it with no need to actually decelerate it. It may be (I haven't done any calculations) that it is infeasible due to cost/deltaV, but I would not rule it out as yet.

Citizen Joe said...

Leading and trailing tethers would be fine, assuming both ship and station were at the same orbit. Lateral tethers/grapples would be better since the orbital speed wouldn't change just the position. Then you could pull yourself back and forth on your orbit by taking on cargo to the north and south as needed to get back on designated flight paths. So you're looking for a wing man position not a train style formation.

Citizen Joe said...

Re: L-Point mass driver. You want the oscillations minimized to capture incoming stuff, but when launching, you want to fire it off at maximum relative velocity and use the launch reaction to nullify most of the oscillation.

Jim Baerg said...

"Don't forget about the fact that you don't pull the ship to the station you pull both together. Large ships could literally pull the station out of orbit with those tactics."

Sure & when you reel the ship out the orbital change is reversed.

Only if more cargo is going one way than the other will the changes add up & some sort of propulsion need to be done on the space station. If more cargo is going down than up, that will (partially) compensate for air drag.

Electrodynamic tether propulsion might be used to compensate for air drag & these cargo transfer effects. If that technology turns out to be impractical, we can attach an ion drive to the space station.

Rick said...

I don't know if there's a formal definition of LEO, but a station a few hundred km up, like the ISS, is readily accessible to surface shuttles, and should have no serious problem with stationkeeping. The proportion of drag cross section to Mass should be comparable to the ISS.

If you are slinging baggage at several km/s, things get a bit more complicated, but no insuperably so, especially given the techlevel for big mass drivers.

The biggest operating challenge, I would think, is catching the baggage - we are not talking 'tens' of meters per second! The mass drivers should not have an onboard crew, because sooner or later an incoming load is going to be off center and deliver a kiloton range whack.

I'm confused about the Langrange points - I thought the unstable ones, L1-3, were just that - staying there is a balancing act, and any perturbation causes you to drift away. The stable points, L4 and L5, I assume have an 'escape velocity.' Any lower relative velocity and you oscillate; if you exceed it you drift away into some other orbit.

But I've never been a big fan of the Lagrange points. Stable points in the middle of nowhere are still in the middle of nowhere. I expect stations to congregate in places where we actually want to go.

Persistent osculation in space is another interesting topic, but probably for another post if not a different blog. ;-)

Citizen Joe said...

Lagrange points deal with the stable orbits around two large bodies.

L1 is between the two bodies. L2 is beyond the smaller body. L3 is 180 is directly opposite the smaller body on the other side of the larger body at the same orbit. L1-3 are all gravitational saddle points. If you move forward or back along the orbital path, gravity wants to pull you to the middle. And it won't continuously oscillate. The momentum gets consumed by the large bodies eventually. As a side note, close passes by planets allow the use of the Oberth maneuver which could almost triple the effect of reaction mass expended in the fly by. That is part of the same reason. Anyway, if you move perpendicular to the orbital path, you'll start to fall away.

L4 and L5 are large areas leading and trailing the smaller body by about 60 degrees. There's an infinitesimally small pull off those points from the two large masses, but typically objects in those areas are more attracted to each other than falling away. Those two L points are like plateaus of gravitic potential. So operations within that large expanse can be done without too much concern of falling out of orbit.

Calsir said...

@Citizen Joe.
In order to use lateral tethers (I will not consider oblique ones :) ), you need to put 2 vessels at the very same altitude on different orbital planes, with more or less the same orbital anomaly: due to orbital mechanics, that is a collision course, as the two vehicles _will_ collide when they reach the orbital intersection points. As for the other configurations, they will necessarily create some girations due to gravity gradient. So far, tethering has been tried only with objects on the same plane at different altitude.

there are stable orbits around the unstable lagrangian points :), like the lissajous orbits (check wikipedia). They require very little station keeping. Also, unstable lagrangian points are much easier to reach than the stable ones, since the instability can be exploited to get there for free.

L4 and L5, on the other hand, are bad for a couple of reasons.
1) they are more expensive to reach, since a vessel must enter the stability region by burning fuel to make its velocity "right".
2) They are in the middle of nowhere and may be full of debris :).
L1 is not in the middle of nowhere, it is close to the planet and it's a good place to put a station.

At the faculty, for a class, we were given the mission to put navigation beacons in the solar system. We ruled out Jovian L4,5 due to the Trojans and Achaeans and the deltaV cost to reach them. The best stable placements were in L1 of Jupiter and Mars. In the end, the mission was proven infeasible, as we would have needed something like the good ole' Energia launcher and a nuclear reactor to power the radio :\

Rick said...

Thanks to both Citizen Joe and Calzir for discussion of Lagrange points, which I've never paid close attention to. (And it shows.)

My bias against Lagrange points is more a matter of seeing other locations as more useful. Especially low orbit around planets to minimize surface lift. No place we can go to at all is anywhere as bad as Earth, but all of them require specialized vehicles, probably powered by chemfuel, making delta v costly.

One important exception: Deimos, especially if it has volatiles that we can get at. If it does, it will likely be the gas station of the inner system.

Calsir - I'm curious, what were the beacons for? Or was the exercise limited to putting them in place? My impression is that even with fairly modest instruments, you can locate yourself in the Solar System to remarkable precision, and don't really need beacon for general navigation. Though presumably spacecraft will have beacons for rendezvous and docking. (Hmm, but does the ISS have anything like beacons?)

Calsir said...

"My bias against Lagrange points is more a matter of seeing other locations as more useful. "

Yes, but L*O does not sound nearly as cool as Lagrange point or L1 :D.

About ISS. According to the orbiter simulator, it has docking beacons. I have no better "source" ;). But it makes sense to put beacons on space stations.

About the beacons we designed: they should have been the interplanetary equivalents of the GPS. The idea was to let other probes to get rid of their locating instrumentation. The first idea was to send 2 of the probes to Jovian L345, so as to cover Sol as a whole with 3 satellites. In the end, we scrapped the whole project as "unfeasible" with current technology.

The main issue was power. The powers-that-be (the teachers) wanted omni-directional radio capability. Trouble, in order to power such a radio we would have needed a nuclear power plant (like those that you can find in Provence). So they relaxed the specs and asked us to provide in-ecliptic navigation; that would have allowed the usage of RTGs, or rather RSG (stirling version). That would have required a spinning satellite with an high gain antenna to transmit the navigational data and another high gain antenna pointed inertially towards Earth (or the other way around). The trouble was that mission specs asked for a 15 years in-station duration, and bearings have the inopportune tendency of wear away much sooner: in facts, when we presented the project at ESTEC, Esa engineers exclaimed "wtf?!" at the mission duration :).

Still, it was an interesting exercise: it took 17 people about 3 months to complete, from conception to scrapping. We went into the 2003 ESA Aurora student design contest, but only won a special prize. I was in mission analysis.

Citizen Joe said...

I'm not sure why you needed radio beacons. There have been numerous sites that explain that "There ain't no stealth in space." And how it is really hard to dump your waste heat. So all you really need is to get a warm body out there and you're done. Maybe spin them so that the heat source flashes at regular intervals. This is a tried and true concept called a Lighthouse. If you need super accurate positioning, you can use a short range omnidirectional 'pinger' once you get close. You really only need the radio transmission for communications. For navigational beacons, it would be the responsibility of ships to detect them. The beacon's responsibility would be to stay in its 'fixed' relative position and produce a unique signature.

Citizen Joe said...

I thought I'd bring up another issue with docking in space. In order to need a docking station in orbit, there needs to be enough space travel to justify it. That travel is pretty much going to need some sort of nuclear powered rocket, whether fusion or fission. The ships are likely to have some sort of radiation shield, but use of the rocket near the station is going to dump all kinds of radiation into the orbit of the station.

I was thinking that the interplanetary drives, the nuclear rockets, would get you into a parking orbit where the radiation won't be a hazard to the station or shipping. Then tugs would come along and move you into position or transfer your load. The tugs might be some sort of closed nuclear power plant that uses steam (water) as propellant. This could be a relatively high thrust engine with low specific impulse. The spent steam would remain in orbit and be collected like dew on morning leaves and that way recycled. This is particularly useful out past the ice line where there are lots of icy asteroids to be used for propellant.

I'm not sure if a steam rocket would be sufficient for planetary lift off. Of course, for planetary lift off (from Earth anyway) there's an atmosphere that you can use for propellant.

Rick said...

Beacons - There's no stealth in space, but especially over interplanetary distances this is implicitly assuming military-grade scan suites. Having said that, we seem to achieve impressive precision now, e.g. getting Huygens onto Titan. So I don't think 'lighthouses' with interplanetary range will be needed.

But presumably we'll put up GPS type constellations around the moon, Mars, and generally any place where people are moving around regularly on the surface. And these can be used by spacecraft in orbital space, if say they are on the opposite side of the planet from the station they plan to make rendezvous with.

Citizen Joe - Ouch, ouch ouch! You hit on a MAJOR issue with rendezvous operations involving nuclear spacecraft. (Except aneutronic fusion, which is a long ways off.) With no atmosphere to attenuate them, neutron flux from any nuke plant has an enormous deadly radius in space. The ship is protected by a shadow shield, but all round shielding would be an excessive mass penalty. Fission plants could be a hazard anywhere near the ship even when shut down.

Cargo transfer can be done by robotic craft. Passenger transfer will be done very delicately.

For the inner system, out to the asteroid belt, solar electric could be an alternative, at least for civil transport. But solar electrics also won't be docking directly to stations, given their huge and delicate solar wings. At least they don't need to stay a thousand km away!

Anonymous said...

Citizen Joe - "The spent steam would remain in orbit and be collected like dew on morning leaves and that way recycled."

Very very fast moving dew that has frozen into fine ice crystals. Assuming it's still concentrated enough to be worth collecting, wouldn't this orbital blizzard juts ablate anything that got in its way?


Rick said...

I think it would disperse far to much to be either capturable or a problem. Exhaust plumes in space disperse really rapidly!

Citizen Joe said...

I think that Tritium batteries can produce about the same amount of power per kilogram mass that solar systems can somewhere a bit beyond earth orbit. So, if you're intentionally using the decayed tritium to form helium-3 then there's that advantage too.

Rick said...

Tritium batteries?? What are those? Powered by the heat of tritium decay?

Calsir said...

The system was for probes and space vehicles. And yes, I know that it is not too hard to navigate sol. It was a very long shot, but you have to remember: students' work is free ;). Besides, it was just an alpha-stage mission plan.

Rick said...

Testing out an idea with cheap labor - free, in fact!

Citizen Joe said...

Rick said...


Anonymous said...

"I'm not sure why you needed radio beacons. There have been numerous sites that explain that "There ain't no stealth in space." And how it is really hard to dump your waste heat. So all you really need is to get a warm body out there and you're done. "

The radio beacons are used for precision close approach, i.e. docking.


Rick said...

Apparently this design exercise was for interplanetary range beacons, not short range docking beacons (which seem pretty inevitable).

Calsir said...

Yep, the idea was to use satellites as lighthouses. To answer Citizen Joe, the idea was that radio transmitters can give information just like the GPS dows nowadays. A single radio transmitter can give both range (through time difference) and direction (if using high gain antennae or radiogoniometres) or simply range from a point.

With the latter, you need to see 4, non co-planar transmitters; with the former, you need to see just one transmitter, since it provides direction and distance to a known point.

I am not saying that it was feasible, I said that we were assigned to do it.

Citizen Joe said...

Polaris then Mars L3,4,5.

Or even cheaper. Polaris, Jupiter, 2 Jupiter Trojan point asteroids. Done...

* quietly hands off calculations to mathematician *

Docking requires FAR more accuracy than what you can get from a GPS system. You just need to get close enough that your own active sensors and LIDAR can reach the dock. The lighthouses don't even need the local area GPS broadcast capability. What they really need are a set of antennae that point at each other and back to home base, like Earth. That way they can talk back and forth to know where they are relative to each other. Then the ship pings Earth for that data to update their system.

Time dilation may be a bigger issue. Presumably, ships will be moving pretty fast so they'll have time slippage and chronometers won't sync up.

Jim Baerg said...

Would any sort of navigation beacon really be needed for interplanetary travel? A good clock, an ephemeris of planetary positions & a device for precisely measuring the angle between 'stars' would do the job.

First locate 2 or 3 'fixed' stars to orient your self.
Now determine the direction a planet or the sun. You know you are somewhere on the half line from the planet in the opposite direction.
Do this with a few more planets & you know you are where these lines come close to intersecting & just how far they are from intersecting gives you the accuracy of your position.

Citizen Joe said...

Until we get high endurance high thrust engines, being able to tell where you are may not help you anyway. All those calculations would have been done at launch. Knowing where you are, and where you destination is, won't help much if it will be on the other side of the Sun when you reach its orbit. You have to think more like someone on a raft or a balloonist. You don't aim for your destination, you maneuver into currents that will carry you there. In this case, its about getting into the same orbit at the same time as your destination.

Jim Baerg said...

Until we get those high performance engines it will be crucial to locate yourself precisely so you know if you are deviating from the orbit needed to get to your destination & make corrections before the deviation gets too big to be corrected with the available propellant.

Rick said...

We must have procedures for doing this now, because midcourse correction burns are common in interplanetary missions, and insertion maneuvers and the like must have a precision requirement within a few hundred or even a few tens of km.

I've never really seen a discussion of how we currently achieve that sort of navigational precision!

Citizen Joe said...

That would require such a high level of nerdity that the Earth itself would slow down as everyone simultaneously reached for their remote controls. It would only be noted that it is difficult to the extreme and they'd show some sort of chalkboard filled with formulae that probably have nothing to do with orbital mechanics.

I don't think it is so much mid course corrections as course corrections once you're far enough away from a gravity well that its effects are negligible. The math gets ridiculously complicated with more than 3 bodies. I think intercepts are also calculated from the destination back to the source. At various steps along the calculation you have windows that you need to hit. Once you pass a step, you aim for the next window. It is like golf. You don't aim for a hole in one, you aim for par.

Calsir said...

It is not about getting on the same orbit (what does orbit mean in this context?) as your target, but to get at the same point (less an offset to avoid collisions) at the same time.

"I don't think it is so much mid course corrections as course corrections once you're far enough away from a gravity well that its effects are negligible."

Actually, the farthest you are from the target the better it is to perform the manoeuvre. Problem: it is also more sensible to error.

"I think intercepts are also calculated from the destination back to the source. "

Wrong. At first approximation (which is quite good for sol) it's just the solution of Lambert's Problem (wiki's got an article about it). Of course, it is a first approximation, since the n-body problem has not a closed solution, but that's pretty good for mission design.

"At various steps along the calculation you have windows that you need to hit. Once you pass a step, you aim for the next window. It is like golf. You don't aim for a hole in one, you aim for par."

With par 2. The "windows" concept does not seem too sound. I don't care that I hit a 1000 sqkm window in the middle of nowhere. I need to get to a point (any point) synched with planet P.

I will keep predicting my position (with patched conics, numerical integration, whatever) and compute the closest encounter with target.

If distance is above a certain tolerance, I manoeuvre to compensate.

When I get close to planet's SOI, I perform a manoeuvre to select the orbital parameters (within achievable limits) that I wish.

Woah. It looks disconnected... It is also 3.00 in the morning here :)

Calsir said...

I would also like to express my deepest apology to English Grammar :)

Rick said...

That would require such a high level of nerdity that the Earth itself would slow down as everyone simultaneously reached for their remote controls.


I think you are right about the analogy to golf (God help us). It is a matter of successive refinements. Which immediately puts me in mind of the wonderful jump navigation sequences in Heinlein's Starman Jones - in balance my fave of his work. The tech is funky, with an enlisted guy using a book of tables to convert computer I/O back and forth between binary and base 10 (!), but it has exactly the idea of successive corrections and refinements of the trajectory.

Rick said...

Calsir - Oops, cross-posted. English grammar has no problem that registers with me. The technical aspects are pretty much above my pay grade, but it reads really cool!

And rather fits the YouTube footage I was just watching, of the first Moon landing.

Kedamono said...

I have a question. These ships aren't exploration probes, taking years to reach their target, they are at the least VASIMR and going like a relative bat out of Hell in space. At some point they will have to flip and decelerate to make that intercept with the planet/space station/asteroid.

This brings up the point that if a ship doesn't flip and start decelerating as it approaches a planet, it becomes a target at some point, not ship.

I figure that there is some wiggle room for the deceleration phase, depending on whether the ship is a big burn and coaster or a small constant burner.

The big burn an coast ship needs only to flip and decell when it's relatively close to the planet. A constant burner has a distinct "midpoint" where it has to flip and decell.

OBS Hard SF: This is something every SciFi show and movie, except one, maybe two, have violated. Every space ship pops out of warp/hyperspace/cruises in from the planet next door and parks itself in orbit like a car parking in a garage.

The two movies? Destination Moon and 2010.

In the first, Heinlein is at the helm and makes damn sure that the ship flips. The second they at least acknowledge that the "Soviet" ship is going to fast and has to aerobrake in Jupiter's atmosphere.

Citizen Joe said...

By reaching the same orbit, I mean being at the same orbital distance and at the same orbital velocity. You could relatively easily get to the same location of a planet at the same time it is there, but that also means you either impact or blast by at ridiculous speed.

As to the flip, planets aren't stationary on a straight line. They are moving targets at different orbits and moving at different speeds. Earth has an orbital speed of about 30 km/sec. Mars has an orbital speed of about 24 km/sec. If you push off from Earth gently, your own orbital speed (centripetal force) will slowly accelerate you outwards as solar gravity decreases with distance. There is also the 75 million km difference between Earth and Mars orbits. If you don't time it right, there's also the difference of the position on the orbital path to make up. You don't just point your ship at a planet and hit the thrusters.

That is actually where the L-points become useful. If you have to wait for a window to open, you can park your ship at an L-point. That also makes L-points useful as refueling stops.

In any case, the VASIMR doesn't produce a lot of thrust (compared to 1 gee) even in its low gear configuration. So, the course would be less flip and more parabolic roll. And you wouldn't notice because it would be so gradual.

When you get into 1G continuous drives (torchships) all of those weird orbital mechanics calculations go out the airlock because you can simply muscle through them. In those cases, yes, you basically point at the planet and flip midway.

Re: FTL and popping in next to a planet. I explained that by redefining the relationship between space/time and mass. Mass is the property of matter that lets it interact with space/time. The other forces deal with how matter reacts with other matter. So large bodies actually distort and drag space/time around with them. So the space around the Sun is being dragged along with the Sun. The space around the Earth is being dragged along by the Earth (and the Sun). So, my FTL method involves shedding your mass and dropping into another dimension where your speed is controlled by the distortions in space time that large bodies make. When you drop back into real space, you pop back in within a large body's space distortion. Thus you're moving at the same relative velocity as that large body. One of the side effects is the conservation of momentum. When you 'shed your mass' you also shed your momentum and it has to go someplace. It ends up creating a shockwave through space, eventually getting absorbed by any matter in its path. When you return to realspace, another shockwave extends out from the ship in the same manner. There are safety features, like the gravitic distortions forcing you out of this other dimension if you get too close to a planet. And you can't really see where you're going, just sense the distortions. So the typical policy is to aim for a gravitic null between a planet and its star or a planet and its moon, and exit there.

Rick said...

I don't think even torchships would do a true brachistochrone orbit with a midway flipover. (Unless, as in Heinlein, it was unsafe to shut down a torch in midflight because they were tricky to relight.)

I'm too lazy to do the math, but accelerating all the way to midpoint* is wasteful of fuel - you'll put on speed that you almost immediately take back off, burning gas for a very small reduction in travel time.

* Midpoint in dynamic terms, allowing for orbits of the departure and arrival planets, etc.

I still don't see any special value to the L-points. If you have to wait for a window, you'll have to wait for it anyway, and you may as well twiddle thumbs where the theaters and bars are.

To first approximation, torchships can just bull right through and treat space as flat. Of course actual navigation will number crunch orbits.

On FTL, this sounds like as good a brand of mumbo jumbo as any, but I'll quote myself from the Tough Guide: So far as genuine scientific plausibility goes, a ship's FTL Drive might just as well be a pretty woman in a white dress who lights some candles and flips tarot cards while chanting in Welsh.

Citizen Joe said...

And thanks to AI's and fancy holograms, it can be :)

Calsir said...

Personally, I love the Alderson Drive. Of course it kills relativity, but I think it's the least important victim (especially since it is not that evident at first glance.

About the brachistochrone, I have never thought about the wastefulness of the acceleration close to mid-point, but I think you are essentially correct. And I am lazier than you are :).

As for L-points. It takes energy to reach one, and energy to leave one. If you have to leave from a parking orbit you'd better stay there, rather than go to, say, L1 and wait there. In any case, you have to expend the energy to reach hyperbolic orbit, so you want to use that energy effectively.

Citizen Joe said...

You could cheat and use the incoming ship's momentum to launch an outgoing ship. That's the idea behind the mass driver.

As to the constant acceleration at midpoint. The previously mentioned Uranus Class refinery ships used a bunch of D-T fusion thrusters, plus at least that many more tugs also with D-T fusion thrusters to perform an Oberth maneuver around Uranus for the launch to Mars orbit (where multiple cyclers would ferry stuff back to Earth). I think they ended up burning about half their reaction mass (ice) in that initial launch. After that, the tugs kicked away from the ship and orbited back to base. The D-D breeder reactors were fired up to refine the stored heavy water (D2O) into Tritium and Helium-3. That process produced a LOT of energy which was used for the VASIMR style thrusters. Then it was six years in space working the refinery and closing the distance to Mars. By the time the D-T thrusters were fired up for the braking thrust, the ship was down to about a tenth of its launch mass, so the onboard D-T thusters were sufficient for braking. The Helium-3 was shipped out and most of the Tritium was pumped into the station's holding tanks for the belters to use as fuel for their betavoltaics (which also decays to helium-3). The refinery ship (now much lighter) could fire up its FTL for a quick hop back to Uranus.

Rick said...

Momentum balancing of mass driver traffic can be done more or less in any orbit, so it still isn't any special advantage for the Lagrange points.

On the refinery ships, if you're producing thrust as a byproduct of an industrial process the wastefulness of brachistochrone orbits does not apply - there's no reason to shut down the refinery process just because it won't shorten travel time by much!

Citizen Joe said...

There are two reasons for the mass drive placement at an L point. One is that orbiting a planet makes the windows of use very short, thus adding another layer of difficulty. The other is that the L-points are relatively stable. The stability is important so that you can move around a bit to catch the incoming vessel (or jink out of the way if it red lines). That moving about isn't going to plummet you into the planet. I think that one of the problems with the L-points inside of Mars is that the other planets perturb those points. So solar Lagrange points 3, 4, 5 are not terribly useful until Mars orbit. Sol-Earth L1 and L2 are still stable enough for some sort of station keeping, like the solar studying satellite in lissajous at L1. L2 is always eclipsed, but is useful for studying the far side of the moon. It may even be a good hiding spot for ships during solar storms.

Planetary Lagrange points have their own uses too. So the Earth Moon L points may be useful for such lings as a communication constellation to stay in touch with the moon. L2 in particular would be important to contact the far side of the moon. L1 would be a good place for a Lunar Space Elevator, assuming we find something there worth exporting.

I don't think the Martian moons are large enough to have suitable L-points, but the Sol-Mars L1 would be a nice spot for some sort of artificial solar wind deflector. Mars doesn't have a magnetic field so its atmosphere gets stripped away faster than we can build it up. It would also be a good place to put large mirrors to reflect sunlight and heat to the polar icecaps of Mars, melting the CO2 there and helping to beef up the atmosphere.

Sol-Mars L3,4,5 could be useful for way stations for asteroid miners. If you miss your window for Mars, you're S.O.L. for like 2-3 years or more. With the additional way stations, a missed window means a 4 or 8 month wait instead. Those additional spots mean more windows for ships coming from the outer planets as well. Incoming cargo would then get shipped to/from earth via a cycler system.

But again, torchships ignore all that stuff.

RE: the refinery ships and thrusters. Their primary thrust comes from the D-T fusion rockets and the tugs. Those are used, and the effect magnified, during the Oberth maneuver around Uranus. The VASIMRs operating during the 6 year trip serve more towards fine tuning the trajectory than adding significant thrust. That is still significantly faster than the Hohmann orbit of 16+ years.

Rick said...

I grant L2 as a repeater station for the lunar farside, and L1 if lunar traffic is heavy enough to call for rail service. But I don't see how the Earth-Moon Lagrange points have any bearing on interplanetary windows. Those are essentially the same wherever you are in Earth-Moon space.

Nor do I see much role for 'way stations' for the asteroid belt, etc., unless they are near a place you might want to go anyway. For regular commercial travel, if you miss a window, tough tahooties. But the rhythm of windows will drive much of economic and social life.

In an emergency ... well most emergencies are over, one way or the other, in less than the time scale of interplanetary travel. And if any emergency response is viable, you can use steep orbits, because fuel economy is not your concern.