Starship Artemis Contract & Lunar Starship

[OTV Booster]

Just to note what I think are a couple of big differences from quadcopter: the copter has extreme levels of acceleration available very quickly; it also has a much lower moment of intertia - tilting over by 20deg in an instant, where required; through widely spaced rotors, it can apply high torque.

Yep, a quadcopter pilot would think the HLS handles like a pig, like a T-38 pilot in an A380 or a speedboat driver in an oil tanker. The pilot will need to re-train, but they share certain fundamentals.

A quad copter controller doesn't have pedals for obvious reasons. IMO chopper controls are more instinctive but personal prejudice is embedded in that opinion. Still, every pilot knows rudder pedals deep in their bones. No reflexes to unlearn and redirect.

[sdsds]

Isaacman says he has asked the astronaut office for plausible scenarios where a pilot taking manual control would be better than aborting the landing. He's missing the point. When there's no possible way to survive, humans want to die while attempting the impossible. Given what we ask of astronauts (and the nature of those we select for the role) is it reasonable to deny them that?

[KilroySmith]

Yes, if it adds development time, weight and complexity.

If adding manual controls resets development, it’ll add years to first flight.   Depending on the complexity of the manual controls, it could add anywhere from 10 to 500 kilos to the ship.   And now you have to deal with failures in the manual control system even when not being used that may affect the mission.

[StarSailor]

Yes, if it adds development time, weight and complexity.

If adding manual controls resets development, it’ll add years to first flight.   Depending on the complexity of the manual controls, it could add anywhere from 10 to 500 kilos to the ship.   And now you have to deal with failures in the manual control system even when not being used that may affect the mission.

Why would it reset development?  Manual controls have been a requirement since before spacex made their bid.  Not a sinlgle person inside SpaceX has claimed HLS will have no manual controls.  In fact it would have been asinine for SpaceX to make a bid on a contract which mandates explicitly that the vessel MUST have manual controls and then spend six years developing it without manual controls.

Here is my boring simple explanation.  HLS has manual controls. HLS has always had manual controls.  HLS will launch with manual controls.  HLS manual controls with will be tested in artemis 3.  HLS manual controls will be available on artemis 4 regardless of if they are used or not.  SpaceX and NASA simply have a disagreement on the exact implementation and/or layout/operation of said manual controls.

[TheRadicalModerate]

Isaacman says he has asked the astronaut office for plausible scenarios where a pilot taking manual control would be better than aborting the landing. He's missing the point. When there's no possible way to survive, humans want to die while attempting the impossible. Given what we ask of astronauts (and the nature of those we select for the role) is it reasonable to deny them that?

If there's almost no chance of survival, you want to be aborting, not landing.  Any manual control mode should reduce the pFail for the abort.  And yes, it's unreasonable to allow them to migrate further into the failure tree to make them feel better--especially if the extra complexity increases overall pLOC and slips the schedule.

The reason that a Moon landing is even marginally within the risk envelope is because abort is incredibly simple:  just increase thrust, stay upright, and then do some (heavily computationally-assisted) orbital mechanics once you've cleared the surface.  Doing anything other than that isn't rational.

This is not to say that there aren't manual override situations that can lead to a successful landing.  But I'm pretty sure that all of them are going to be oriented toward making a last-second decision that the current landing site is worse than expected, and there's one a couple hundred yards downrange that looks pretty good.  Anything more complicated than that?  Abort.

[meekGee]

Isaacman says he has asked the astronaut office for plausible scenarios where a pilot taking manual control would be better than aborting the landing. He's missing the point. When there's no possible way to survive, humans want to die while attempting the impossible. Given what we ask of astronauts (and the nature of those we select for the role) is it reasonable to deny them that?

If there's almost no chance of survival, you want to be aborting, not landing.  Any manual control mode should reduce the pFail for the abort.  And yes, it's unreasonable to allow them to migrate further into the failure tree to make them feel better--especially if the extra complexity increases overall pLOC and slips the schedule.

The reason that a Moon landing is even marginally within the risk envelope is because abort is incredibly simple:  just increase thrust, stay upright, and then do some (heavily computationally-assisted) orbital mechanics once you've cleared the surface.  Doing anything other than that isn't rational.

This is not to say that there aren't manual override situations that can lead to a successful landing.  But I'm pretty sure that all of them are going to be oriented toward making a last-second decision that the current landing site is worse than expected, and there's one a couple hundred yards downrange that looks pretty good.  Anything more complicated than that?  Abort.

That's exactly right.  The manual control you need is a button that says "I've lost confidence in the landing system (be it computer, propulsion, sensing, etc) and it's best we try to to abort with what we still have.

Isaacman is showing he's a discerning party here - he's out his finger on the exact absurdity.

[KilroySmith]

Why would it reset development?  Manual controls have been a requirement since before spacex made their bid.  Not a sinlgle person inside SpaceX has claimed HLS will have no manual controls.  In fact it would have been asinine for SpaceX to make a bid on a contract which mandates explicitly that the vessel MUST have manual controls and then spend six years developing it without manual controls.

Here is my boring simple explanation.  HLS has manual controls. HLS has always had manual controls.  HLS will launch with manual controls.  HLS manual controls with will be tested in artemis 3.  HLS manual controls will be available on artemis 4 regardless of if they are used or not.  SpaceX and NASA simply have a disagreement on the exact implementation and/or layout/operation of said manual controls.

I 100% agree with you.  My point is simply that if, to NASA, "manual controls" means add helicopter-style controls usable by a suited astronaut, and have multiple manual modes ranging from "Put'er down over there instead" to rapid control of individual thrusters, there's a huge amount of work to do that would likely invalidate many of the currently architected, reviewed, designed, reviewed, implemented, reviewed, tested, and reviewed systems. That's not a quick thing to recover from.

[Bob Shaw]

Pilots gotta pilot, so we know what they'll say. A big red physical ABORT button is a great idea, but the idea of a steely-eyed missileman making a better landing attempt than an automated system is just a romantic fantasy.

[Paul451]

Train in a chopper to start for that seat of the pants thing then move to the simulator for a more representative feel.

Oh hell no. Helicopters are the worst aircraft to develop instinctual rocket-VTOL experience, due to the complex interconnection between the gyroscopic effects of the rotor and tail. You described the nominal actions of each control, but in a helicopter they don't actually work that way: if you want to climb, you don't just add throttle, you perform a complex dance of throttle/rudder/collective/cyclic to avoid the whole thing spinning out of control the moment you touch any one control.

The only reason I would allow a helo pilot anywhere near a rocket lander is because they can usually "turn off" the instincts they had to learn to fly helicopters.

[sdsds]

When there's no possible way to survive, humans want to die while attempting the impossible.

If there's almost no chance of survival, you want to be aborting, not landing.

Consider a horrible mishap situation during final descent where the automated system concludes that landing would be unsurvivable and also that abort can't get back to Orion rendezvous. Maybe for sake of discussion it's an major MMOD strike on a propellant tank. An autopilot might choose to abort to an orbit where Orion can't reach it. Once there, it loses all propulsion and thereby dooms the crew to slow death. A human pilot might choose to slam into the lunar surface, going out in a quick blaze of glory. Should we not at least give them a second 'big red button' to inhibit abort? Do we not trust them to choose their own fate?

More seriously, fighter jocks might choose to slam the 'inhibit abort' button inappropriately, equivalent to failing to eject due to over-confidence in their ability to land. Might Isaacman be projecting his own personality onto NASA astronauts?

[Paul451]

An autopilot might choose to abort to an orbit where Orion can't reach it. Once there, it loses all propulsion and thereby dooms the crew to slow death. A human pilot might choose to slam into the lunar surface, going out in a quick blaze of glory. [...] Do we not trust them to choose their own fate?

No. Not even a little bit. There's no way for a pilot, or anyone else, in such a split-second decision, to know if there is a viable rescue scenario. Nor should any one crew member be able to make that decision unilaterally on behalf of the other three crew members.

(Plus I doubt there's a scenario where they would be able to so quickly determine that they could or couldn't get to a viable orbit. Training would be, "ship goes bang during landing, twist the abort handle".)

Might Isaacman be projecting his own personality onto NASA astronauts?

I think you might be subconsciously projecting a mythical "blaze of glory" personality onto NASA astronauts.

[meekGee]

When there's no possible way to survive, humans want to die while attempting the impossible.

If there's almost no chance of survival, you want to be aborting, not landing.

Consider a horrible mishap situation during final descent where the automated system concludes that landing would be unsurvivable and also that abort can't get back to Orion rendezvous. Maybe for sake of discussion it's an major MMOD strike on a propellant tank. An autopilot might choose to abort to an orbit where Orion can't reach it. Once there, it loses all propulsion and thereby dooms the crew to slow death. A human pilot might choose to slam into the lunar surface, going out in a quick blaze of glory. Should we not at least give them a second 'big red button' to inhibit abort? Do we not trust them to choose their own fate?

More seriously, fighter jocks might choose to slam the 'inhibit abort' button inappropriately, equivalent to failing to eject due to over-confidence in their ability to land. Might Isaacman be projecting his own personality onto NASA astronauts?

Where did this "choose your form of death" come from?  This is a design of a transport system, not a self serve gallows.

[mikelepage]

Isaacman says he has asked the astronaut office for plausible scenarios where a pilot taking manual control would be better than aborting the landing. He's missing the point. When there's no possible way to survive, humans want to die while attempting the impossible. Given what we ask of astronauts (and the nature of those we select for the role) is it reasonable to deny them that?

I don’t for a moment believe any astronaut would have that kind of a “blaze of glory” mentality, but once you rule that out, I don’t think it’s that much of a stretch to think that there are plausible scenarios where manual control would save lives, it’s just a matter of how likely they are.

Imagining them is a fun exercise. The following assumes that under the nominal plan, HLS lands with just enough prop (plus margin) to make it back to LLO… so how about this:

Suppose HLS is coming in for final landing on the lunar surface, and tanks are pressurising for the “hip” thrusters (no idea if this is how they’ll actually work, but I’m guessing it will be somewhat similar to how the second stage preps before meco).

Then, as the hip landing thrusters start up, something blows up. There’s also a stuck valve or something, and one of the (16?) hip thrusters becomes stuck on, resulting in a small but non-negligible off-axis thrust.

Computer decides to abort the landing and return to orbit, shutting down the hip thrusters and relighting the raptors. But pilot immediately notices prop is being spent much faster than expected for burn - whatever happened means that the prop is leaking somewhere. At this rate HLS will not have enough prop to make orbit.

Importantly, at this point HLS is still only single digit kms/miles from landing site, and not moving very fast, but moving in the wrong direction. Getting back to the prime landing site is beyond the capability of the landing hip thrusters. A manual abort to surface landing is still possible however, and can still put HLS down within range of surface transport.

Pilot takes manual control, shuts down raptors, and restarts hip thrusters. The single stuck valve remains, and the same landing thruster recommences its off axis thrust. But the pilot can compensate for this.

With satellite imagery of the site, and other crew acting as spotters, the pilot successfully puts HLS down at a new site, 5-10km away from the original landing site.

Hopefully at least one of the lunar transport vehicles is already on the surface, so they teleoperate this to shuttle supplies (or drive it themselves) to however much already exists of the moonbase. They stay on the surface for x weeks/months until the next HLS comes to get them. Ticker tape parade for the pilot and crew when they return. 

What are the odds of this or some similarly convoluted scenario happening? Probably fairly low, but if that was my ride, I’d much rather have a pilot onboard trying to figure it out.

[clongton]

An autopilot might choose to abort to an orbit where Orion can't reach it.

There is no Orion. It's docked way back at the Gateway. HLS is on its own. Abort to orbit is the ONLY path open.
Get back to LLO and compute the trajectory back to Gateway.

[sdsds]

[...] Then, as the hip landing thrusters start up, something blows up. There’s also a stuck valve or something, and one of the (16?) hip thrusters becomes stuck on, resulting in a small but non-negligible off-axis thrust.

Computer decides to abort the landing and return to orbit, shutting down the hip thrusters and relighting the raptors. But pilot immediately notices prop is being spent much faster than expected for burn - whatever happened means that the prop is leaking somewhere. At this rate HLS will not have enough prop to make orbit.

Advocates for automation will assert that many milliseconds before the human could notice the propellant deficiency and make a decision, the autopilot would also have noticed it and would already have taken optimal action.

The trap Isaacman has set for the astronaut office isn't very subtle. For each scenario they dream up that the autopilot doesn't handle right Isaacman will say, "Thanks for finding that! We've fixed it. Now are you satisfied?" Software testers from decades ago faced this trap all the time. To prevent buggy software from being released they needed to show that for every bug that was discovered and fixed there were n more still there, lurking undetected. Note those undetected bugs weren't necessarily less harmful; only harder to detect.

Somewhat parenthetically, this is why it makes sense to focus on manual control for unsurvivable situations. I do regret my prior 'blaze of glory' phrasing; that doesn't at all represent my view of the core values of NASA astronauts.

An autopilot might choose to abort to an orbit where Orion can't reach it.

There is no Orion. It's docked way back at the Gateway. HLS is on its own. Abort to orbit is the ONLY path open.
Get back to LLO and compute the trajectory back to Gateway.

Sure, LLO makes a fine intermediate trajectory. Computing the optimal trajectory back to Orion (wherever it is located) can happen in the first few milliseconds after detecting the abort condition though. Unless it somehow depends on those ultra-slow carbon-based compute units.

[OTV Booster]

Train in a chopper to start for that seat of the pants thing then move to the simulator for a more representative feel.

Oh hell no. Helicopters are the worst aircraft to develop instinctual rocket-VTOL experience, due to the complex interconnection between the gyroscopic effects of the rotor and tail. You described the nominal actions of each control, but in a helicopter they don't actually work that way: if you want to climb, you don't just add throttle, you perform a complex dance of throttle/rudder/collective/cyclic to avoid the whole thing spinning out of control the moment you touch any one control.

The only reason I would allow a helo pilot anywhere near a rocket lander is because they can usually "turn off" the instincts they had to learn to fly helicopters.

I haven't been in a chopper for over 50 years. Before my time, when pulling pitch on the collective, throttle had to be rolled in manually. When I was in the game throttle was handled by a governor but counter torque was still manual.


Wow. Counter torque is still manual? I am gobsmacked. Sounds like there's some money to be made there.

[MickQ]

Train in a chopper to start for that seat of the pants thing then move to the simulator for a more representative feel.

Oh hell no. Helicopters are the worst aircraft to develop instinctual rocket-VTOL experience, due to the complex interconnection between the gyroscopic effects of the rotor and tail. You described the nominal actions of each control, but in a helicopter they don't actually work that way: if you want to climb, you don't just add throttle, you perform a complex dance of throttle/rudder/collective/cyclic to avoid the whole thing spinning out of control the moment you touch any one control.

The only reason I would allow a helo pilot anywhere near a rocket lander is because they can usually "turn off" the instincts they had to learn to fly helicopters.

I haven't been in a chopper for over 50 years. Before my time, when pulling pitch on the collective, throttle had to be rolled in manually. When I was in the game throttle was handled by a governor but counter torque was still manual.


Wow. Counter torque is still manual? I am gobsmacked. Sounds like there's some money to be made there.

They could always try a contra rotating helo.  Negligible gyroscopic effects.

[InterestedEngineer]

  The onboard pilot can feel the lurch and already knows which cluster of engines to take readings on - and if good, really good, might have already taken the first steps towards compensating before even being able to verbalize the problem.

Ever flow airplane on instruments?

your feelings lie to use.  Seat of the pants crashes your plane.

[mikelepage]

Advocates for automation will assert that many milliseconds before the human could notice the propellant deficiency and make a decision, the autopilot would also have noticed it and would already have taken optimal action.

Of course it will notice it faster. My point is, there will always be some convoluted series of events that will place the scenario outside the range of scenarios for which the autopilot even knows what the optimal action is. Unknown unknowns, etc.

I guess this is the same old argument about why to take humans at all, vs doing robotic-only exploration. Humans are creative in ways that machines are not (AI slop is generative, not creative). Ideally, we should automate as much as we can, then leave the rest of the choices to the people… as long as the cost of doing so isn’t unacceptable new failure modes.

I will find AI so much more trustworthy once it finally demonstrates that it knows when to say “I don’t know”.

[thespacecow]

Imagining them is a fun exercise. The following assumes that under the nominal plan, HLS lands with just enough prop (plus margin) to make it back to LLO… so how about this:

<snip>

What are the odds of this or some similarly convoluted scenario happening? Probably fairly low, but if that was my ride, I’d much rather have a pilot onboard trying to figure it out.

I don't get why this scenario requires manual control of ship orientation/descent rate/etc, let alone individual thrusters.

Seems to me what you want is just a way to communicate human's intent to the flight computer: Stop abort to orbit, attempt a landing asap. This can be as easy as a button says "Land NOW", similar to Crew Dragon's "Deorbit NOW" button, and some logic to override abort to orbit when this is pressed. There's no need for any more complicated manual control, once flight computer knows the intent, it can counteract the off axis thrust and attempt a landing much better than any human can.