Humanoid Robots for Moon and Mars

[edzieba]

This preparatory work (delivery, unpackaging, installation and possibly more) is where the "interacting with existing physical infrastructure in existing built environments" occurs, because once it is installed it tends to largely just do its task.

Since Mars is a Tabula Rasa, if you design all your packaging and installation sequences to require humanoid and only humanoid intervention, you've failed at the first hurdle. A look inside factories that deal with high levels of automation will show just how much packaging and setup machines can accomplish if you design that packaging and setup with machines in mind.

Remember, a human is NOT the optimum universal task-performer architecture. It's just the shape we're used to, so we've shaped our tasks to be performable by humans. Once you start adding tasks beyond human capability, this fallacy becomes clear.

[SpaceLizard]

Remember, a human is NOT the optimum universal task-performer architecture.

Please provide a better universal task-performer architecture. Because I have a hard time believing any wheel based architecture, or handless architecture will be better suited for any environment that hasn't been specifically designed for its benefit. Maybe AI could conjure up some sort of octopus derived or liquid metal terminator concept that would be more optimal for universal tasks, but those are much farther away technologically than a 'mere' humanoid robot...

[Twark_Main]

One of the comments that got my attention was the difficulty of unloading the vehicles and setting up various devices from there. Conclusion was that Optimus robots would be required. My reaction was to wonder if that person was aware of current forklift capabilities including some autonomous capabilities.

This applies to pretty much all "use a robot to operate machinery that a human can also operate" tasks: the trend Earthside is already to move humans off of the machinery and remotely operate them (sometimes from a tethered controller nearby, sometimes from a wireless controller nearby, often from a networked controller at an arbitrary distance away). In this situation, sending heavy machinery to Mars with operator cabs so they can be dual-operated by humans or humanoid robots is the short of humorously bass-ackwards thinking that looks cool in hollywood, but has been eclipsed by actual progress: if you want to automate you machinery you automated it by eliminating the physical control step, not putting a robot with robot hands at the controls. Nobody builds self-driving card by putting a robot in the driver's seat of an existing car. Nobody builds UAVs by putting a humanoid robot in the cockpit.

Yes, thank you.

I hope this didn't need to be said, but I'm glad you put it so eloquently.

In general, humanoid robots fill the specific niche of minimising the cost of interacting with existing human-optimised physical infrastructure in existing built environments. When these conditions are not met, robotics optimised for the task have always been the better option - cheaper, more reliable, more efficient.
 
On Mars, there is no bult environment, and there is no existing human-operated equipment. The niche that humanoid robots serve does not exist - and it will only exist if you make design and architecture choices that are actively worse than solutions we already know work well.

This is misunderstanding why humans are on job sites. Yes, some are in operator cabs. But the relevant humans are the ones doing "long tail" tasks: clearing debris out of that finicky tool attachment mechanism, grabbing that dust shield part that came detached so maintenance can bolt it back on, smacking the actuator that gets stuck 0.3% of the time even after repeated actuation cycles.

Wheels are cheaper than legs (especially maintenance), so you just have 1-2 Optimus wrapped up in a "pod" somewhere on the vehicle, suitably dressed in dust-proof bunny suits. When needed they deploy, walk over and do the task, then re-stow on the vehicle.

All the advantages of a vehicle, all the flexibility of an Optimus (while not wasting wear-and-tear trekking kilometers across the desert).

Don't think "operator cab," think "stowed accessory repair bot." Basically R2-D2.

[clongton]

Nobody builds self-driving (card) cars by putting a robot in the driver's seat of an existing car.

Edit mine
True, but you might put a humanoid robot in the driver's seat of a normal car    (ducking now  )

[Coastal Ron]

robotics optimised for the task have always been the better option - cheaper, more reliable, more efficient.
 
On Mars, there is no bult environment, and there is no existing human-operated equipment. The niche that humanoid robots serve does not exist - and it will only exist if you make design and architecture choices that are actively worse than solutions we already know work well.

Sure for the task alone, but how many of the specialised robots on Earth deliver themselves to their work location and do all the unpackaging and installation work on their own?

As of today though, humanoid robots would NOT be able to perform anywhere close to all of the requirements needed to unpack and install new equipment. And in order to maximize the amount of unpacking and installation a humanoid robot could do, you would have to do what edzieba theorized, which is make design and architecture choices that are less efficient, cost more, require more material, etc.

Maybe FSD cars are getting close but even then I can argue there is preparatory work of building roads, plus signage, traffic light equipment etc. Road building certainly uses specialised equipment which could be automated at least to some extent but humans are also needed, it isn't all fully automated.

The vast amount of work done in road building is by machines, with humans guiding them.

Tesla FSD is a great example of where something relatively simple (i.e. navigating in 2D) is not so simple for modern robotic systems. And you want to stick them in a harsh 3D environment far from human controllers? 

You are, it seems, arguing for 20 specialised machines to do 20-40 different preparatory tasks (plus the specialised equipment). Then if one of the 20 breaks down, you have a problem. 20 Optimus robots each able to adapt and/or be trained to do hundreds of different things, if one breaks down then you carry on with 19.

You are over-generalizing what a humanoid robot can do, because I doubt they would be able to 100% do any of what those 20-40 different equipments could do.

So early on, 20 Optimus robots (plus the specialised equipment) is the way you go. Later on when there are thousands of robots and machines then more specialisation for better efficiency becomes possible.

Have you ever observed an Optimus robot perform any complex task here on Earth, in a remote location? Because until Tesla demonstrates that ability HERE ON EARTH, why would you think it can be done on Mars? When will Tesla be able to do that?

[Twark_Main]

...

Have you ever observed an Optimus robot perform any complex task here on Earth, in a remote location? Because until Tesla demonstrates that ability HERE ON EARTH, why would you think it can be done on Mars? When will Tesla be able to do that?

#NotHowProgressWorks

Nobody has a crystal ball of course, but Tesla has the best talent, a good approach, and an efficient "startup" organization structure. They also (uniquely) have the right mix of AI, hardware design, and mass manufacturing. All three are necessary.

Many people have bet against Elon Musk in the past. Accelerating AI progress is now his primary focus. Short (or scoff) at your own peril.   

[crandles57]

Have you ever observed an Optimus robot perform any complex task here on Earth, in a remote location? Because until Tesla demonstrates that ability HERE ON EARTH, why would you think it can be done on Mars? When will Tesla be able to do that?

Is Optimus ready to be sold yet? Clearly not. Is it advancing rapidly in what they have it doing? Yes.

Will it be ready to ship and perfect in 1 year, even allowing updates 6 months later after landing? I seriously doubt it, but then I doubt there will be anything landed on mars next synod certainly not with serious cargo.

Will it be ready and perfect in 3 years, allowing updates 6 months later (after landing)? Perfect? No. Substantial progress in this 3.5 years? Definitely. Will some things take a few attempts with improvements in the instructions or AI model between attempts such that things move slowly forward? Yes. Is this efficient at all or even any of the tasks? No.

Is this better than taking 5 times as long as this 3.5 year period to design, build, train and test your 20 specialist robots to do 40 or so different tasks and then after this 20 year period you can efficiently do these 40 tasks but then one breaks and leaves you stuck. What do you want, some slow inefficient progress or even slower for design time, nicely efficient but more vulnerable to failures?

Re vulnerable to failures: Suppose you find none of your 20 specialist robots are suitable for the 41st task that you haven't got around to thinking about? Maybe some combination of your different robots can be trained to work together to fix the task better than a few of the Optimus working together? Perhaps, but for a quick fix are we more used to thinking about how humans would work together to do a task and would this lead to a quicker fix? I am not at all sure I am justified on this point.

I am sure I do want to see it starting soon rather than waiting 20 years for the design, build, train and test your 20 specialist robots. That lots of design and development work is going into Optimus to make it adaptable to hundreds of tasks on Earth in order to sell millions of them here means that this design and development work if not completely free comes at a very reduced price, perhaps mainly some development work to adapt to lower g environment. (Note that it is much easier to stand up on mars as you have longer to shift your mass before you fall over but I assume you probably still need some adaptations to stop Optimus over-reacting to small movements that it doesn't need to react to and to allow larger angle of lean before taking steps and so on.)

[Coastal Ron]

...
Have you ever observed an Optimus robot perform any complex task here on Earth, in a remote location? Because until Tesla demonstrates that ability HERE ON EARTH, why would you think it can be done on Mars? When will Tesla be able to do that?

Nobody has a crystal ball of course, but Tesla has the best talent, a good approach, and an efficient "startup" organization structure. They also (uniquely) have the right mix of AI, hardware design, and mass manufacturing.

I don't think anyone is in a position to judge any of those qualities.

And while there are many metaphors that could be used, the one I'll use for today is that what you are saying above could have been said about the Wright Brothers years after they made their first flight. Of course what we can see through the lens of history is that while achieving human flight turned out to be fairly easy after we discovered the basics, perfecting aircraft usefulness took years of iteration and improvements in many fields.

I think we are in the same situation with humanoid robots. I've been following this field for 50 years, and while lots of progress has been made in the past decade, there are still some areas (like end effectors) that have made little progress when compared to what human hands and feet can do.

Many people have bet against Elon Musk in the past.

I'm not betting against anyone, just like I never bet against any private space companies.

I'm just providing perspective as a future possible user, and explaining why what is available today is too limited for the price likely to be paid.

And Exhibit A today is from a party that Tesla threw, where they controlled all aspects of public interaction with the Optimus robot, and this still happened: 
https://www.reddit.com/r/teslamotors/comments/1pfzxt0/tesla_optimus/

[Coastal Ron]

Have you ever observed an Optimus robot perform any complex task here on Earth, in a remote location? Because until Tesla demonstrates that ability HERE ON EARTH, why would you think it can be done on Mars? When will Tesla be able to do that?

Is Optimus ready to be sold yet? Clearly not.

They don't have to sell it at this point, just "install" them into actual environments for work. Don't you wonder why they haven't yet?

Is it advancing rapidly in what they have it doing? Yes.

Eh, not so much. What is your metric for progress? How do you measure it. For instance, what is the goal that you are measuring against?

Will it be ready to ship and perfect in 1 year, even allowing updates 6 months later after landing? I seriously doubt it, but then I doubt there will be anything landed on mars next synod certainly not with serious cargo.

While software will always be something that can use updates, the biggest issues I see with the current generation of humanoid robots is hardware related, which are locked in when they are built. In other words, if you can't show that it can do something here on Earth as soon as it walks out of the factory, then it likely won't be able to do it off of Earth.

Will it be ready and perfect in 3 years, allowing updates 6 months later (after landing)? Perfect? No. Substantial progress in this 3.5 years? Definitely.

Where will the progress come from? What do you think is the area that needs the most work?

Is this better than taking 5 times as long as this 3.5 year period to design, build, train and test your 20 specialist robots to do 40 or so different tasks and then after this 20 year period you can efficiently do these 40 tasks but then one breaks and leaves you stuck. What do you want, some slow inefficient progress or even slower for design time, nicely efficient but more vulnerable to failures?

Without specifics it is hard to understand if anything you are saying is likely. I mean I can make up reasons all day long for why something won't work, but none of those reasons may be based in reality. Which is why I keep asking for real life examples, because otherwise we're arguing about how many angels can fit on the head of pin. 

For instance, how many 40kg boxes can Optimus move 100m in a day? To know this we'd have to know the carrying capacity of Optimus, the shape of the boxes, the amount of energy required for Optimus to make one trip, the battery capacity of Optimus, and the charging rate for Optimus.

What do you think the answers are to this simple task?

[lamontagne]

Have you ever observed an Optimus robot perform any complex task here on Earth, in a remote location? Because until Tesla demonstrates that ability HERE ON EARTH, why would you think it can be done on Mars? When will Tesla be able to do that?

Is Optimus ready to be sold yet? Clearly not.

They don't have to sell it at this point, just "install" them into actual environments for work. Don't you wonder why they haven't yet?

Is it advancing rapidly in what they have it doing? Yes.

Eh, not so much. What is your metric for progress? How do you measure it. For instance, what is the goal that you are measuring against?

Will it be ready to ship and perfect in 1 year, even allowing updates 6 months later after landing? I seriously doubt it, but then I doubt there will be anything landed on mars next synod certainly not with serious cargo.

While software will always be something that can use updates, the biggest issues I see with the current generation of humanoid robots is hardware related, which are locked in when they are built. In other words, if you can't show that it can do something here on Earth as soon as it walks out of the factory, then it likely won't be able to do it off of Earth.

Will it be ready and perfect in 3 years, allowing updates 6 months later (after landing)? Perfect? No. Substantial progress in this 3.5 years? Definitely.

Where will the progress come from? What do you think is the area that needs the most work?

Is this better than taking 5 times as long as this 3.5 year period to design, build, train and test your 20 specialist robots to do 40 or so different tasks and then after this 20 year period you can efficiently do these 40 tasks but then one breaks and leaves you stuck. What do you want, some slow inefficient progress or even slower for design time, nicely efficient but more vulnerable to failures?

Without specifics it is hard to understand if anything you are saying is likely. I mean I can make up reasons all day long for why something won't work, but none of those reasons may be based in reality. Which is why I keep asking for real life examples, because otherwise we're arguing about how many angels can fit on the head of pin. 

For instance, how many 40kg boxes can Optimus move 100m in a day? To know this we'd have to know the carrying capacity of Optimus, the shape of the boxes, the amount of energy required for Optimus to make one trip, the battery capacity of Optimus, and the charging rate for Optimus.

What do you think the answers are to this simple task?

According to a direct Google search, 100 W idle, 500 W active and performing tasks, 2,3 kWh battery pack.
Recharging off a dedicated 240 volts 15 amp plug should take 45min to 1 hour.

So presuming it can lift 40 kg, and walks at 5000 m per hour, let's choose a distance of 100m, that's 50 boxes per hour.  There has to be some mnipulation time, so let's divide by two. 25 boxes per hour, x 40 kg = 1 tonne per hour.  about 20 hours per day if you remove charging times, so 20 tonnes of boxes per day per robot for a typical 100m path.

[lamontagne]

One of the difficulties with lights out factories, that more or less killed them up till now, is the difficulty or rearranging a production line if changes need to be made, and how production came to a halt if any problem came along.  This should be easier with robots. 
Boring and repetitive taks are hard for humans.  But not for robots.
Robots don't die.  They malfunction, or stop.  That 'x' task was not a good plan and a human died.  Sad.  Inquiery. Safety investigation. Legal disputes. Perhaps major loss of production.  The same task failure and a robot was destroyed?  Too bad.
Robots eat electricity.  Humans eat food.  Food will be very difficult to produce on Mars, and vey heavy to bring along.  About one tonne per year per person.  The solar power required to 'feed' a robot for a year?  Not so much.  Perhaps 100 kg, at most.

[etudiant]

Lamontagne shows that even current day robots can perform useful work. Future versions will undoubtedly be better.
The MTBF of these robots is a critical aspect.
Lamontagne envisions a duty cycle of 20 hr/day, apart from maintenance and charging time.
I've never seen any robot perform for a full hour as yet, much less humanoid robots.
That suggests that a step function improvement in reliability is a fundamental requirement for robot assisted work to happen.
Does anyone have any insights on this aspect?

[Cheapchips]

I've never seen any robot perform for a full hour as yet, much less humanoid robots.

You can watch a few 7 hour video of a team of Agility's Digit robots doing box moving, if that's your cup of tea.

https://www.youtube.com/live/cMMTtP-a6uo?si=i5oWyxXtKMove7vz

[crandles57]

Without specifics it is hard to understand if anything you are saying is likely. Which is why I keep asking for real life examples, because otherwise we're arguing about how many angels can fit on the head of pin. 

Exactly, which is why I am waiting for you to provide details of your 20? specialised robots and how they are going to tackle the jobs I suggested earlier in the post below. Then details of why this is better or worse than sending 20 Optimus robots.

So the first starship with useful cargo lands. You don't want heavy equipment flying around on launch or even minor course corrections so heavy equipment is strapped in with straps and buckles while other things are held in place with lighter staps and a variety of packing, clips, velcro etc. The first thing you need is power ....

So are you going to design all the solar arrays to be able to extricate themselves from where they are stored/secured in starship and from any packaging and be able to move themselves to the lift so they can be lowered to the ground then move themselves to a suitable location, unfurl their solar array and plug in any cables. Why have all the necessary appendages to do all that on each and every one of the solar arrays when some Optimus robots can be used to do this? Are the Optimus going to be brilliant and get everything perfect on first attempt without any human supervision? Of course not.

You activate a couple at a time and watch how they get on, providing feedback as needed. When there is enough power you can activate a few more and soon we are finished with this task for a while and we want to move on to something else.

Something else mightfor an example be scouting local and then more distant areas. Not really sure what the ground penetrating radar equipment will be like. Maybe handheld for Optimus to carry? Not being very sure of this perhaps your version is more like a self propelled wheeled cart so that you don't need an optimus to carry or push it along. That may well make a lot more sense than adding motorised wheels to each and every solar array. Do I imagine it having FSD like AI for route planning and extra appendages to free itself from how it is secured in the Starship? Err no. I do imagine if it is self propelled for there to be fairly simple electronics to follow a given route which human planners provide after seeing images from Optimus scouting.

I just don't see how unpackaging and unloading and probably several other preparatory tasks is going to get sensibly done. Just adding all sorts of appendages to everything particularly things that only need to be placed in position once or only rarely moved seems quite silly to me and obviously over-engineered and a waste of mass that has to be transported to Mars.

A fork lift easily beats Optimus on speed and efficiency of moving boxes over level or levelish ground with few obstacles. I don't care, I want the adaptability of Optimus. Give the forklift a different task of moving gravel or maybe those 40kg mass boxes are more like cylinder shaped roll out solar arrays which the robots not only have to place in position but also roll out the array and plug in some cables.

[MickQ]

Or Optimus could load half a dozen boxes onto a wheeled cart and push it to the required location, unload and return for more  🤔

[redneck]

A fork lift easily beats Optimus on speed and efficiency of moving boxes over level or levelish ground with few obstacles. I don't care, I want the adaptability of Optimus. Give the forklift a different task of moving gravel or maybe those 40kg mass boxes are more like cylinder shaped roll out solar arrays which the robots not only have to place in position but also roll out the array and plug in some cables.

Rough terrain forklifts routinely use a variety of attachments to move gravel, cylinders, boxes, and other machinery. The ones I'm familiar with operate on ground that is quite difficult for humans to move efficiently on. Soft sand and steep slopes among other things.

[crandles57]

Or Optimus could load half a dozen boxes onto a wheeled cart and push it to the required location, unload and return for more  🤔

Perhaps even more so if it is push cart to location A where item A is unloaded, push cart to location B where item B is unloaded through to item J then return to load another 10 boxes which is more like the solar array lay out problem.

So what is best?
i) Optimus plus lightweight cart or
ii) Forklift plus Optimus for unloading unrolling and plugging in or
iii) something else. If so what?

Forklift sounds like it eats mass budget of transport to Mars to me but maybe I am generally curious about the early stages and later on it becomes more viable if we keep finding tasks the forklift can be adapted to.

[Coastal Ron]

One of the difficulties with lights out factories, that more or less killed them up till now, is the difficulty or rearranging a production line if changes need to be made, and how production came to a halt if any problem came along.  This should be easier with robots.

I've spent decades working in factories, both large and small. The small ones would be the hardest for a humanoid robot to work in because of the number of individual tasks that have to be done in such a small area. In other words, there is lots of setup and tear-down events requiring (usually) high levels of precision.

Boring and repetitive taks are hard for humans.

What? Are you kidding? Throughout human history one of the things that humans have excelled at are monotonous jobs - repetitive tasks.

And in the human world it is the humans doing the monotonous tasks that usually FIND the problems first, and then sometimes they are actually able to use their human brains and experience to fix the problems. This is especially true in smaller factories that have a wide variety of products that they build. That has also been one of the challenges with human labor, is that they accumulate what we used to called "tribal knowledge" about how things are done. That is something that humanoid robots could capture, but first they have to learn it, and learning it requires logic that today's humanoid robots DO NOT HAVE.

Remember today's humanoid robots don't "think". They are not "reasoning machines". They are being worked on, but I think that will end up being closer to what people call AGI (Artificial General Intelligence), and that is still so far into the future that we don't know when it will truly come.

But not for robots.

The same reasoning and calculations used today to determine if humans should be replaced with dedicated machines will be used to decide whether humanoid robots should be replaced with dedicated machines.

Robots don't die.  They malfunction, or stop.  That 'x' task was not a good plan and a human died.  Sad.  Inquiery. Safety investigation. Legal disputes. Perhaps major loss of production.  The same task failure and a robot was destroyed?  Too bad.

Um, in Elon Musk's version of the future, there won't be OSHA on Mars... 

Seriously though, today it is pretty rare when humans get injured on the job. I never saw anyone die at the places I worked at, nor did we ever need an ambulance.

Robots eat electricity.  Humans eat food.  Food will be very difficult to produce on Mars, and vey heavy to bring along.  About one tonne per year per person.  The solar power required to 'feed' a robot for a year?  Not so much.  Perhaps 100 kg, at most.

For what will likely be a short period of time on Mars, there won't be humans. But the whole point of sending humanoid robots to Mars is to prepare for a horde of free human labor to show up. So if you're going to feed those colonists, you might as well have them work for their food (and air, and...  ).

[Coastal Ron]

Without specifics it is hard to understand if anything you are saying is likely. Which is why I keep asking for real life examples, because otherwise we're arguing about how many angels can fit on the head of pin. 

Exactly, which is why I am waiting for you to provide details of your 20? specialised robots and how they are going to tackle the jobs I suggested earlier in the post below. Then details of why this is better or worse than sending 20 Optimus robots.

You're the one that came up with that number, how did you arrive at such a figure?

Remember I'm the one saying that as of today there isn't much a humanoid robot WOULD be able to do. And my proof of that is to point out the complete lack of any humanoid robots deployed doing actual day to day work in remote locations, unsupervised and unsupported by humans. Do you know of any?

A fork lift easily beats Optimus on speed and efficiency of moving boxes over level or levelish ground with few obstacles. I don't care, I want the adaptability of Optimus. Give the forklift a different task of moving gravel or maybe those 40kg mass boxes are more like cylinder shaped roll out solar arrays which the robots not only have to place in position but also roll out the array and plug in some cables.

There will be a whole host of restrictions that will determine which actual solutions are tried out. Power requirements are one, and wheeled vehicles are much more power efficient than walking humanoids.

But for doing small tasks, where hand-like manipulation is required, even then that might be solved with a robot that looks like a Centaur (horse-human hybrid) - though maybe it is a humanoid robot riding around in an electric cart.

This is why you first need to define what the "job to be done", and what the constraints are, before you can settle on a solution.

I think humanoid robots should be sent to Mars for testing, I'm just not sure if everyone will want to depend on them immediately.

[Paul451]

Forklift sounds like it eats mass budget of transport to Mars to me

Articulated loaders (not just forklifts) come in every scale you could want.