Humanoid Robots for Moon and Mars

[Coastal Ron]

No, it doesn't. You keep saying that "cost" is the reason why Elon Musk doesn't let his Optimus robots do work humans can do, but that is not only wrong, it is NOT the reason.

Why? Because Tesla is building Optimus robots anyways, without selling them, so why wouldn't they place them in situations where they can demonstrate what they can do? Or do you expect them to just sit around in crates?

I am trying to say cost is the reason that other humanoid robots that have existed for a while are not around out in the workplace doing stuff is because doing something slower than a human and needing 1:1 supervision makes no financial sense to deploy this.

The flaw in your logic is that Optimus won't be able to learn how to do something UNLESS it does something. UNLESS it has that 1:1 human being with it initially until the performance of Optimus is validated and understood.

This is EXACTLY what Tesla is doing with Full Self Driving (FSD), and what every autonomous transportation company is doing. They are not sitting there thinking "oh, my autonomous car is too expensive right now", and keeping it parked. No, they are out there USING it to learn.

Optimus is designed for cost efficient large scale manufacturing.

This is a PR line you have been fed (which you are repeating), and I can say that being someone that has actually worked in high volume production factories. And you can actually see how Optimus is being made today in Tesla videos, and it ain't "high volume".

Optimus, version 3 of the hardware is only just? not yet? rolling off the production line in volume.

Why are they building them? Who are they building them for?

Look, I was a factory scheduling manager, and sometimes you build a few of something to test out your production processes, but that is ONLY because you plan to ramp up as quickly as possible to full production. Meaning you have a distribution channel set up or customers identified.

Who is Tesla selling Optimus to? And they would only do that if they have validated the Product/Market Fit of their design - which I don't think they have done yet.

...but Musk seems to be talking up rapid ramp up in production volumes suggesting that version 3 hardware is good enough (maybe wrongly like FSD hardware?).

Musk says many things - Elon Musk has his own "reality distortion field" (RDF), and on NSF we fondly talk about "Elon Time" when trying to interpret when something will actually happen versus when he says it will. 

Anyway I am trying to say there is plenty of adequate reasons why no Optimus sales yet...

Occam's razor would indicate that they don't yet provide enough value for anyone to use them. And this is not just an Elon-Optimus issue, it is the same problem globally for everyone developing humanoid robots.

[crandles57]

The flaw in your logic is that Optimus won't be able to learn how to do something UNLESS it does something. UNLESS it has that 1:1 human being with it initially until the performance of Optimus is validated and understood.

This is EXACTLY what Tesla is doing with Full Self Driving (FSD), and what every autonomous transportation company is doing. They are not sitting there thinking "oh, my autonomous car is too expensive right now", and keeping it parked. No, they are out there USING it to learn.

Re 1:1 human being

A couple of things
https://x.com/Tesla_Optimus/status/1925047336256078302

This video suggests they can use videos of humans doing things.

Then there is Sim2Real
https://www.notateslaapp.com/news/2732/tesla-engineers-reveal-how-optimus-learns-and-show-off-its-dance-moves-video

Sim-to-Real Learning
The recurring theme from the Optimus team is the power of sim-to-real transfer using reinforcement learning. This approach involves training the AI model extensively in a simulated virtual environment, where it can learn complex behaviors (like walking, balancing, and now dancing) through trial and error at an accelerated pace. The best part is that it can all be done without risking damage to the physical hardware and done across multiple nodes of Tesla hardware. Imagine thousands of Optimus bots learning to dance all at once - except virtually.

Once the AI masters these skills in the simulation, the challenge lies in transferring that learning effectively to the real-world robot - which is the sim-to-real step.

These latest dance demonstrations suggest Tesla is making some fairly substantial strides in bridging that gap, allowing Optimus to translate simulated learning into real-world physical competence.

So between these training methods, I don't think I follow your objection to using videos of human(s) doing something and then have a mars environment simulator allowing thousands of attempts at the task in the simulated mars environment until you get a reasonable model of what the optimus robot needs to be aware of and to do to complete the task. Then upload that to an Optimus robot on Mars. They may well have extra steps in the process like first putting it to an Earth simulator to develop the AI model then give it to an actual Optimus robot on Earth to see how it gets on. Maybe they can tweak things if it is going in the wrong direction to steer it more towards the solution they want?

Some tasks may well be harder to convert from Earth to Mars environment, hopefully many others won't be so hard. Using body mass to be more effective with a pick-axe may well be difficult one. This might cause them to work around the problem e.g. resort to using jackhammer more often.

Yes more training is needed and will continue to be needed with new tasks. What is the problem with doing in simulators on Earth but which simulate Mars environment?

Lots of local optimums that are not the global optimum can make the simulators job of finding the global optimum a lot harder requiring far more numerous simulations to be done but if this isn't much of a problem for Earth, will it be much worse of a problem for Mars?

[Vultur]

While there are no humans on Mars then there is no such human alternative and it is a matter of best method that works. Slow and inefficient but eventually getting job done beats a method that either fails or involves huge long waits to design something more efficient.

How long will there be significant payloads on Mars but no humans, though? By the most recent plans made public it looks like maybe ~4 years (2 synods). Maybe ~2 years (1 synod) if the first uncrewed synod is largely testing without delivering significant payloads.

I think I will be impressed if it turns out to be less than 4 synods but I could easily be wrong.

Why? If the necessary hardware is successfully landed, why *not* send humans in the next synod?

If it is 4+ synods, is there a high chance that it tends to morph into almost entirely robotic? Will Musk prevent that happening?

Depends on if the goal changes.

The goal as stated is human civilization on Mars, doing everything with robots is not a step toward that.

But SpaceX seems to be taking a turn toward an AI focus (which I am not comfortable with, I don't want the now very likely AI bubble burst to take them down too).

[crandles57]

Why? If the necessary hardware is successfully landed, why *not* send humans in the next synod?

Mainly what is necessary to bring the humans back. First thought starts at we need to demonstrate propellant production so the equipment for small scale test is sent. However then you start thinking of more requirement to make it safe enough to send humans and you start shifting the requirements and start thinking you need to:

* find sufficient recoverable ice that can be mined
* set up Earth pressure habitat and safe zone for solar storms
* not only produce the propellants in volume but possibly also transfer them to the return starship and be able to prep that starship for launch

Edit: when you have found sufficient recoverable ice then you start thinking have we demonstrated ice mining at sufficient scale to be sure that that won't be a problem.

I could easily be completely wrong.

[Vultur]

Why? If the necessary hardware is successfully landed, why *not* send humans in the next synod?

Mainly what is necessary to bring the humans back. First thought starts at we need to demonstrate propellant production so the equipment for small scale test is sent. However then you start thinking of more requirement to make it safe enough to send humans and you start shifting the requirements and start thinking you need to:

* find sufficient recoverable ice that can be mined
* set up Earth pressure habitat and safe zone for solar storms
* not only produce the propellants in volume but possibly also transfer them to the return starship and be able to prep that starship for launch

Edit: when you have found sufficient recoverable ice then you start thinking have we demonstrated ice mining at sufficient scale to be sure that that won't be a problem.

I could easily be completely wrong.

While the plan could have changed, the plan as publicly stated so far is to land the ISRU equipment before humans but have humans set it up.

I think a lot of work has already been done on ice location.

The Starships are suitable as habitats for quite a while, and solar storms are not really an issue on Mars - the atmosphere does provide significant shielding against that.

As for "safe enough" if the goal is settlement then waiting another synod or two before return is possible is not a deal breaker. Just bring extra supplies.

[Cheapchips]

While there are no humans on Mars then there is no such human alternative and it is a matter of best method that works. Slow and inefficient but eventually getting job done beats a method that either fails or involves huge long waits to design something more efficient.

How long will there be significant payloads on Mars but no humans, though? By the most recent plans made public it looks like maybe ~4 years (2 synods). Maybe ~2 years (1 synod) if the first uncrewed synod is largely testing without delivering significant payloads.

If you assume a useful humanoid / general purpose robot, you'll always be sending more robots that people. 

Humans need tons of food and equipment along with a significant kWh per day power budget.  Humanoid robots need some area of solar production and that's it.  Maybe some spare parts or you just recycle within the fleet until it's topped up at the next synod.

Robots are significantly more useful with humans present. Novel tasks without humans either don't happen or you have to sim train them on Earth and upload the policy.  Local humans can just jump in and teleoperate.  It saves an EVA and adds to the training data. 

The important thing is that jump in teleoperation is a low occupancy activity for the human. They can jump in and out of robots performing at multiple locations / tasks.

Say a widget needs swapping out.  The robot could retrieve it, go to the location of the broke widget. Human jumps in to remove the broken widget and fit the new one. Robot takes the broken widget to the repair shop. Significant time saver for the human, they only did a tiny bit of work. They could have done another task with another robot in the same time window. Or just had a lovely cup of tea.

That degree of autonomy + teleoperation combo is pretty much there today.

1X's next year of beta participation will be interesting to watch. Their launch model is partly autonomous, partly teleoperation when needed.

[Vultur]

While there are no humans on Mars then there is no such human alternative and it is a matter of best method that works. Slow and inefficient but eventually getting job done beats a method that either fails or involves huge long waits to design something more efficient.

How long will there be significant payloads on Mars but no humans, though? By the most recent plans made public it looks like maybe ~4 years (2 synods). Maybe ~2 years (1 synod) if the first uncrewed synod is largely testing without delivering significant payloads.

If you assume a useful humanoid / general purpose robot, you'll always be sending more robots that people. 

I think this depends on both goals and scale. If the goal is to build a human *civilization* on Mars (not merely an outpost) you need jobs other than overseeing robots.

A small number of humans with teleoperation/ semi autonomous but overseen robots as force multiplier in the early stages of buildout, sure. Once things are more built out and the goal is building a civilization not a survival/refueling infrastructure... That might change.

As for scale:

Humans need tons of food and equipment along with a significant kWh per day power budget. 

Once things scale up food will be produced locally (minus some low mass items like coffee, spices, etc). Mass of equipment per person is probably also a scale thing - a huge pressurized volume (eg lava tube) is much more forgiving and doesn't require individual homes to be pressure vessels with independent ECLSS.

Robots are significantly more useful with humans present. Novel tasks without humans either don't happen or you have to sim train them on Earth and upload the policy.  Local humans can just jump in and teleoperate.  It saves an EVA and adds to the training data. 

The important thing is that jump in teleoperation is a low occupancy activity for the human. They can jump in and out of robots performing at multiple locations / tasks.

Oh, teleoperation would absolutely be used, no questions there. But that's way less demanding than robots operating autonomously at a tens of minutes light lag.

(I still think that use of *humanoid* robots like Optimus is an "if you have a hammer" thing though. Teleoperation doesn't require humanoid forms, and something with fewer joints is probably optimal for work outside on Mars in dust conditions )

[Cheapchips]

Teleoperation doesn't require humanoid forms, and something with fewer joints is probably optimal for work outside on Mars in dust conditions

Goods (local) teleoperation needs a waist proxy and a small footprint, alongside arms and hands.  It's why the humanoid form is pretty good.  Sunday's Memo shows off a good way to do a wheeled base.

[clongton]

Musk says many things - Elon Musk has his own "reality distortion field" (RDF), and on NSF we fondly talk about "Elon Time" when trying to interpret when something will actually happen versus when he says it will. 

Agree. Mr. Musk is a visionary, which is a good thing, but more often than not he makes statements about progress with his end/intermediate goal in mind as if all the intermediate steps to get there have or are about to be completed (even though they're not done yet). He doesn't "sweat the details"; vetting, hiring and then working the best design/build engineers he can find to "sweat the details" for him. A lot of us here come from the industry, where this kind of "leaving the details to the staff" is not usually practiced. Most managers bury themselves in the weeds - our weeds -, which is why most projects seem to take longer than they should. That's not Mr. Musk's style. His style is to set the goals for his staff, get them to buy into his vision, get a really, REALLY good idea of what they need to pull it off, provide all of it in spades, incentivize them to beat the targets and then get the hell out of the way and let them identify and work thru the details (the weeds). It took quite a while for a lot of us to understand that management style because it's not what we're used to. It's really cool, and scary efficient. But stuff gets done under his watch - really quickly - even if it doesn't jive with Elon Time.

[clongton]

But SpaceX seems to be taking a turn toward an AI focus (which I am not comfortable with, I don't want the now very likely AI bubble burst to take them down too).

I'm not prepared to go that far. In my opinion, Mr. Musk has identified the need for a large work force on the Martian surface and is leaning on AI Robotics to be a force multiplier for what will initially be, a small human worker population. The more workers he can put on the surface, human or robotic, the faster the first colony can take root. It's not going to be an Optimus colony. It's going to be a human colony with robotic help. Human colonists will overtake any robotic workers in numbers soon enough. But without the robots, that would take a much longer time.
AI Bubble burst? Even if that did occur, I doubt it would hurt Mr. Musk's plans.

[crandles57]

While the plan could have changed, the plan as publicly stated so far is to land the ISRU equipment before humans but have humans set it up.

I think a lot of work has already been done on ice location.

The Starships are suitable as habitats for quite a while, and solar storms are not really an issue on Mars - the atmosphere does provide significant shielding against that.

As for "safe enough" if the goal is settlement then waiting another synod or two before return is possible is not a deal breaker. Just bring extra supplies.

Interesting. You may well have found more than me.
Wikipedia

Musk's plans for the first crewed Mars mission state that it will consist of approximately 12 people, with goals to "build and troubleshoot the propellant plant and Mars Base Alpha power system" and establish a "rudimentary base".[citation needed]

So that doesn't seem sure of the source for that. Do you know of such a source? I would like to see it to see if it helps assess how detailed SpaceX plans are and whether we can discern things like whether the build and troubleshoot is for a full size plant but a smaller scale test plant has already been robotically built and tested.

>"if the goal is settlement"

SpaceX has stated that it plans to build a crewed base on Mars which it hopes will grow into a self-sufficient colony

Sounds to me like crewed base first with settlement only coming later. If it is only a crewed based at first then presumably you want the crews to be able to rotate and come back to Earth and the crew probably also want to know it is possible without huge delays.

[Twark_Main]

However then you start thinking of more requirement to make it safe enough to send humans and you start shifting the requirements and start thinking you need to...

Endlessly shift the safety goalposts as soon as the old ones are achieved??   


Sufficiently advanced handwringing can stymie any space mission. It's impossible to reach a point where there's zero risk. Eventually you have to accept the risk and go for it, or else someone bolder will do it for you.

[crandles57]

However then you start thinking of more requirement to make it safe enough to send humans and you start shifting the requirements and start thinking you need to...

Endlessly shift the safety goalposts as soon as the old ones are achieved??   


Sufficiently advanced handwringing can stymie any space mission. It's impossible to reach a point where there's zero risk. Eventually you have to accept the risk and go for it, or else someone bolder will do it for you.

Hopefully not, if the plans are detailed enough and well thought through.

If the detail we have is things like 'Musk said' the first humans would be involved in building propellant plant, is that from a well thought through plan or just something Musk said off the top of his head hoping that is the way it would go and there are no detailed plans just some fuzzy thoughts?

In this only fuzzy thoughts level of plans case then I think you start in crude detail - we must be able to produce propellants in order to get back but then as you flesh out the details of the plans then you start adding stricter requirements.

[Twark_Main]

While the plan could have changed, the plan as publicly stated so far is to land the ISRU equipment before humans but have humans set it up.

I think a lot of work has already been done on ice location.

The Starships are suitable as habitats for quite a while, and solar storms are not really an issue on Mars - the atmosphere does provide significant shielding against that.

As for "safe enough" if the goal is settlement then waiting another synod or two before return is possible is not a deal breaker. Just bring extra supplies.

Interesting. You may well have found more than me.
Wikipedia

Musk's plans for the first crewed Mars mission state that it will consist of approximately 12 people, with goals to "build and troubleshoot the propellant plant and Mars Base Alpha power system" and establish a "rudimentary base".[citation needed]

So that doesn't seem sure of the source for that. Do you know of such a source?

He says it constantly, so I immediately know you're not really paying attention.

Digging up anything Elon Musk related is practically impossible because the (SEO) well is so poisoned.   

I would like to see it to see if it helps assess how detailed SpaceX plans are and whether we can discern things like whether the build and troubleshoot is for a full size plant but a smaller scale test plant has already been robotically built and tested.

It doesn't.

>"if the goal is settlement"

SpaceX has stated that it plans to build a crewed base on Mars which it hopes will grow into a self-sufficient colony

Sounds to me like crewed base first with settlement only coming later.

Potatoh / potatow. This is mere semantics, not a meaningful technical distinction.

If it is only a crewed based at first then presumably you want the crews to be able to rotate and come back to Earth and the crew probably also want to know it is possible without huge delays.

Presumably there are plenty of (less timid) candidates in line behind you.  NEXT! 

[DanClemmensen]

Wikipedia

Musk's plans for the first crewed Mars mission state that it will consist of approximately 12 people, with goals to "build and troubleshoot the propellant plant and Mars Base Alpha power system" and establish a "rudimentary base".[citation needed]

So that doesn't seem sure of the source for that. Do you know of such a source? I would like to see it to see if it helps assess how detailed SpaceX plans are and whether we can discern things like whether the build and troubleshoot is for a full size plant but a smaller scale test plant has already been robotically built and tested.

That's from
     https://en.wikipedia.org/wiki/SpaceX_Mars_colonization_program
Some archeology in the article's history log shows that it was originally sourced to this 2016 Geekwire article:
   https://www.geekwire.com/2016/spacex-elon-musk-geeks-out-mars-reddit/
Subsequent edits to Wikipedia to try to bring it more up to date managed to delete the reference, and it then later got the "citation needed". If you have the time, you can try to bring it up to date, but it's a mish-mash right now.

[Twark_Main]

However then you start thinking of more requirement to make it safe enough to send humans and you start shifting the requirements and start thinking you need to...

Endlessly shift the safety goalposts as soon as the old ones are achieved??   


Sufficiently advanced handwringing can stymie any space mission. It's impossible to reach a point where there's zero risk. Eventually you have to accept the risk and go for it, or else someone bolder will do it for you.

Hopefully not, if the plans are detailed enough and well thought through.

Your hope is in vain.

Sitting in your living room at home is not a "zero risk" activity, so space travel has no hope.

If the detail we have is things like 'Musk said' the first humans would be involved in building propellant plant, is that from a well thought through plan or just something Musk said off the top of his head hoping that is the way it would go and there are no detailed plans just some fuzzy thoughts?

You want "fuzzy thoughts" at this stage, because if you attempt detailed plans they're going to change, so all that work is wasted.

I expect Musk's version of "fuzzy plans" still has more thought behind it than NSF users' version of "well thought out plans."

In this only fuzzy thoughts level of plans case then I think you start in crude detail - we must be able to produce propellants in order to get back but then as you flesh out the details of the plans then you start adding stricter requirements.

You need to start adding more specific requirements. "Stricter" is entirely your decision.

Unfortunately you're deciding in a way that tries to maximize cost if there's even a slight risk. You seem to have no sense of there being (and this is a good thing) a risk-vs-cost tradeoff.

[Twark_Main]

I think this reveals differing attitudes (and comfort levels) toward uncertainty.

Crandles seems to be trying to use Musk Kremlinology to suss out individual decisions. But I expect Musk sees the "shape" of the problem not as a single holy blessed incarnation, but instead as a series of trades. "We we encounter this we'll do this, otherwise we'll switch to Plan B which is this" etc etc.

Even that makes it sound too concrete (like there's one holy blessed flowchart), but you don't want your thinking to be that rigid. Agile thinking is what separates SpaceX from OldSpace.

[crandles57]

You need to start adding more specific requirements. "Stricter" is entirely your decision.

Unfortunately you're deciding in a way that tries to maximize cost if there's even a slight risk. You seem to have no sense of there being (and this is a good thing) a risk-vs-cost tradeoff.

Thanks for the answers (and to Dan)

You are correct 'stricter' is the wrong word. I think it is more that sometimes the "added more specific" requirements are not particularly onerous and do not lead to extra time before you are ready to send humans but sometimes they might do. A mix of no added time and some added time adds up to some added time.

Minimise the risk whatever the effects on cost and delay seems like a way to ensure you are beaten to it. Some risk has to be accepted to get it done faster. So I agree and maybe I am being too risk averse in suggesting all my extra requirements need to be done before sending any humans. Great, if the longer delay before sending humans isn't needed though I tend to think expecting no delay is optimistic and I might remain impressed if it is as short as 2 or 3 synods.

>Agile thinking is what separates SpaceX from OldSpace.
Yep, old space way takes a long time, cough, 60 years to return to the moon if ever?, cough. New space way gives ridiculously short timelines and there are some delays but it gets done faster than old space way. If we all agree on this, that seems like we should expect some delays?

[crandles57]

Draft grokipedia Edit - says it has been approved
https://grokipedia.com/page/SpaceX_ambition_of_colonizing_Mars
Add new subsection under "Phased Mission Architecture" called "Robotic Precursors: Optimus Humanoids"

It has been announced that Tesla's Optimus robot will be sent to Mars before humans. These robots have relevant features including AI neural network controlled providing potential for autonomy and remote operation, human-like dexterity, flexible mobility, and energy efficiency. It has been suggested they could be used in habitat construction, surface exploration and site analysis, solar panel deployment and power maintenance, agricultural automation, repair and maintenance tasks, and radiation monitoring and shielding. However, the extent to which this robotic precursor strategy alters previous SpaceX plans remains unclear. Earlier visions emphasized human-led efforts from the outset, whereas Optimus could potentially accelerate infrastructure development and reduce initial human risk exposure. A broad question of whether Optimus can achieve sufficient reliability for Earth-based applications remains open before being deployed on Mars where more specific challenges like whether the robots will have adequate' durability in the face of Martian dust, which could compromise their functionality over time. These factors introduce uncertainty regarding the timeline and effectiveness of the robotic phase, though they are seen as critical to de-risking human missions.
refs
https://x.com/elonmusk/status/1900774290682683612
https://spacetime24.com/teslas-optimus-on-mars-mission/


Probably lots more that could be added but I think I will leave it there for time being.

[Coastal Ron]

Draft grokipedia Edit - says it has been approved
https://grokipedia.com/page/SpaceX_ambition_of_colonizing_Mars
Add new subsection under "Phased Mission Architecture" called "Robotic Precursors: Optimus Humanoids"

...
A broad question of whether Optimus can achieve sufficient reliability for Earth-based applications remains open before being deployed on Mars...

This is exactly what I have been talking about.

That they need to use Optimus here on Earth first, to validate a whole multitude of things about Optimus and humanoid robots in general. That they haven't been, to any great extent, tells me that Optimus (and the many other humanoid robots around the world) are not yet ready for doing "work" on a daily basis - something that would be required if Optimus is sent to Mars.