Friday, December 27, 2013

Who needs humans?

This is in response to Quantum G's question "Why do humans need to return to the Moon to get resources to make "consumables and propellant", if robots can be sent to do that instead?"

Just let autonomous and/or teleoperated robots do all the work. Who needs humans?

Quantum G should try working in an actual mine. As an ASU student, I spent four summers working in the Phelps Dodge copper mine in Ajo, Arizona. At the top of every bulletin board was Murphy's Law: "What Can Go Wrong, Will."  And that was followed by many variations and corollaries of Murphy's Law.

Unlike a factory floor, mines are an uncontrolled, unpredictable environment. The unexpected can and does happen. When it does, human ingenuity is called for. You cannot write algorithms that anticipate every unforeseen problem.

Not that I'm against robots. See
Puppets, Telerobots & James Cameron,
Surgical Robots, and
Give NASA's SLS money to DARPA.
I believe improved robotics will be a major game changer when it comes to exploitation of space resources.

The moon is more amenable to tele robots than most locations in our solar system. At 384,400 kilometers from earth, light lag latency is only 3 seconds. Since signal strength falls with inverse square of distance, lunar tele robots would enjoy much better bandwidth than machines on remote asteroids or Mars. Good bandwidth is important for immersive tele-presence as well as control of agile, dexterous robots.

And there are technologies that can mitigate a 3 second reaction time. For example Big Dog's balance or Google Car's collision avoidance.

Even so, a multitude of tasks are much easier with constant sensory feedback in real time. Things like finding a dropped hex nut. A 3 second light lag can make normally quick and easy chores time consuming and difficult. Robots controlled by humans in neighboring habs would be much more able than bots controlled from earth's surface.

And then there's the question of maintenance. Who maintains the robots?

Here is an article on mining giant Rio Tinto's "autonomous" robots. These driverless trucks move back and forth along well maintained and predictable routes. And they are closely monitored by nearby humans. Machines in less predictable environments such as the shovels are still human operated. And all the machines, whether "autonomous" or human operated, are maintained by humans.

Mines sans humans are still well beyond the state of art for earthly mines, much less mines in environments where we have zero operating experience.

Robots may reduce the need for human presence. But they won't completely eliminate the need for humans, not for a long while.

There is also important information to be gained from humans on the moon. What gravity do humans need to stay healthy? As I mention in What's the minimum spin hab?, this is still not known. If the moon's 1/6 gravity keeps humans healthy, that makes minimum spin habs for asteroid workers more than six times less massive. It would also indicate humans are okay living with Martian gravity.

22 comments:

QuantumG said...

Thanks. Of course, I agree with you completely. Way back in 2010 it really did seem like NASA was going to pursue a *cheap* robotic return to the Moon. So when (not if) the robots broke, it'd be painless to send another, and another, and another. At some point this becomes less economical than just having humans on-site, but so long as NASA continues to see lunar landers (carrying humans) as costing billions and billions and requiring new heavy lift rocket development that costs billions and billions, then the economy of just-send-another-robot is very long.

Anonymous said...

I have to counter quite a few points here.

Robots controlled by humans in neighboring habs would be much more able than bots controlled from earth's surface.

This is not really supported by many studies looking into human-robot cooperation. There is a notion of "sliding autonomy" and it mostly turns out that robots and humans are most efficient working together when robots have a large degree of autonomy up to a certain decision level.

I.e. modern robots don't need help picking up the nut, but they do need help deciding if it should be done or not. Bandwidth and latency are really minor factors in that kind of assistance.

Mines sans humans are still well beyond the state of art for earthly mines, much less mines in environments where we have zero operating experience.

This is not completely true either. Deep sea or underwater mining is being enabled by ROVs, or remotely operated vehicles. Essentially, they are telerobots. A lot of work underwater happens with ROVs these days. Note that Solwara plant is not yet in operation, but not due technology but the politics of the situation.

The advances in robotics are accelerating, there are a few underlying technological breakthroughs that enable this.

Jon Goff said...

Hop,

I also agree with your take on this. While I run a space-robotics related startup, I think something complex as ISRU is going to really benefit from people on the ground. Especially for maintenance. Sure, you can probably land a purely-roobotic demo system that can prove the concept of ISRU. But once you move even to pilot-plant scales, the ability to have at least a person or two on-site is likely going to be extremely valuable.

If it turns out that humans can survive long-term on 1/6G, it might be possible to send the people there "one-way for a while", where they stay at least until they have the pilot plant running well enough to make their return propellant.

While it's true that if you do lunar landers the NASA way, with HLVs, and LSAM like monstrosities, it will make sending humans ridiculously expensive. Fortunately there are non-NASA approaches that look technically feasible. Raising money to execute on those non-NASA approaches? That's more iffy at the moment.

~Jon

Hollister David said...
This comment has been removed by the author.
Hollister David said...

Please identify yourself, I don't like talking to an "anonymous".

Googling "deep sea mining" I don't see that it exists yet. There is an outfit called Nautilus that plans to do deep sea mining.

I suppose you could call drilling wells a form of mining. ROV's were developed in the 60's. We were reaching remote resource bodies via wells before then.

But it is fair to cite ROV use in off shore drilling. These ROVs are being used to build substantial infrastructure at depths beyond the reach of divers.

But so far as I can tell, these ROVs are linked to human operators by fiber optic cables. The operators seem to operate them in real time.

Can you give some examples of working ROVs with sliding autonomy? Googling I can find autonomous surveillance machines. They're good at looking around and gathering data. But I couldn't find examples of autonomous machines building stuff.

At Seatronics ROV tooling are a variety of tools ROVs employ. Cutting, grinding, using a wrench are the sort of things I imagine an ROV doing while being operated by a human in real time.

Karl Hallowell said...

"Mines sans humans are still well beyond the state of art for earthly mines"

No, robot/teleoperated-only mines just aren't cost effective on Earth when you have people who can easily live next door. If it costs a hundred thousand dollars per kg of human (you have to include mass of life support) to send people, then the economics are a bit different.

roboticist said...

There are multiple names for telerobots - undersea they are called ROVs and various institutes and companies around the world are operating them in various capacities.

In the air they are called "UAVs" and they are operated by various militaries.

Each of these have various degrees of autonomy built in, and almost none are reliant on persistent realtime high bandwidth links to human operators these days.

As for autonomous machines building stuff - if you ever visited any modern factory floor in last couple of decades, you'll see nothing but. And even that industry is going fundamental shift now, as robots are being deployed that are self learning and can work alongside humans without cages around them.

Robotics is a very wide industry area and funnily enough different applications dont even often cross over.

For more on terrestrial telerobotic mining, listen to Greg Baiden here
http://www.youtube.com/watch?v=FVmD3s1DODI

For what completely autonomous robots can do these days, look at these two buddies cooking pancakes ( running fully open source software and operating completely autonomously )
http://www.willowgarage.com/blog/2010/10/21/tum-rosie-and-pr2-james-make-pancakes-together

Anonymous said...

This is a good post about the value of telerobots. But the presumption that humans will be needed in situ because "who fixes the robots?" is somewhat naive. Rio Tinto has humans to fix their telerobots because, well, on Earth, humans are cheap. They won't be on the Moon. Not by a long shot. And task completion isn't a matter of "algorithms". It's a matter of reproducing human perception and dexterity.

No reason why a telerobot can't pick up a nut, and fix another telerobot, if its vision and dexterity, as relayed to the human operator, is good. The human is essentially there. There are many examples of telerobots operating effectively where humans have a hard time operating.

The cost of latency is very significant in space, though, as you note. That can be dramatically reduced not just by putting human controllers in nearby surface habs, but in habs orbiting overhead. Those controllers can operate the telerobots in almost real time. The latter strategy is operationally advantageous, because surface systems for humans aren't required. So I very much agree that latency is a serious penalty, though it is a penalty that human spaceflight can fix.

Just call me "anon" for short.

roboticist said...

No, robot/teleoperated-only mines just aren't cost effective on Earth

Exactly the opposite, and im not going to spend a ton of energy proving otherwise, sufficient if you listen to Greg Baiden talk that i linked to above. Labor costs drive mining costs enormously and there is a push for much more automation.

In fact, labor unions are actually holding the progress back there in many places.

roboticist said...

BTW, i forgot a very large industry segment running very autonomous robots - shipping and ports. Especially maritime shipping.

Look up "automated guided vehicles" or pull up some videos on how advanced shipping ports operate these days - port of Rotterdam for instance.

Hollister David said...

Thanks for a great video. I had thought of Baiden as a power-point miner, an interesting academic whose ideas hadn't seen much use in the world outside academia. That video changed my opinion. I was delighted to see some of his products enjoying actual use and generating revenue.

Baiden seemed okay with lunar light lag latency, I expect he knows what he's talking about.

At one point someone evidently asked a question about lunar dust. He answered dust is part of the mining business, he was used to it. In my opinion he was casually dismissing a serious problem, lunar dust is more abrasive than the dust he's accustomed to.

He also didn't talk much about maintenance. At around 29:50 he talks about excavating a cave on a lunar wall and putting in an airlock. So it seems the lunar mine he envisions would have humans. He stipulated the airlock would be big enough to allow the machines inside. So I'm guessing he expects humans within a hab to maintain the machines.

I was surprised to see the stuff Baiden's machines were doing. However I still believe a lunar mine would need some human presence.

Hollister David said...

"As for autonomous machines building stuff - if you ever visited any modern factory floor in last couple of decades, you'll see nothing but."

As I've already mentioned, factory floors are a controlled, predictable environment. Mines are not.

kert said...

Note that even as ISS was never designed with robotic servicing in mind and is all pretty old hardware, Dextre has successfully performed repair work without astronauts being involved.

And dextre itself is not exactly bleeding edge tech compared to terrestrial robots.

QuantumG said...

Can anyone recommend a good human-level dexterity example of teleoperated robots? The last time I asked this I got a whole bunch of mumbo jumbo about surgical robots. It really should be too much to ask that someone has built a robot that can put together a piece of flat-pack furniture, should it? Even if it has custom tools, that'd be fine.

roboticist said...

As I've already mentioned, factory floors are a controlled, predictable environment.

This was maybe still true 10, 5 years ago, but not so much anymore.

There is a lot of push in for deploying much more flexible robots in uncaged environments, working alongside humans, specifically for low volume applications where dedicated assembly line setup is not cost effective.

See Motoman SDA10 dual arm robots for example designed explicitly for these kinds of apps. Some of these get mounted on mobile bases, too.

Hollister David said...

"Can anyone recommend a good human-level dexterity example of teleoperated robots?"

Trent, at the last Leprecon Science Fiction a doctor who uses surgical robots did a presentation. It was amazing! I wrote about it at Surgical Robots

The "hands" were more like a lobster's pinchers than a hand with 4 fingers and an opposable thumb. But they were still impressive.

QuantumG said...

Hol, I don't think surgical robot dexterity translates to the kind of torque and dexterity you need for real world wrench turning. There might be something relevant in here: http://www.youtube.com/user/DARPAtv/videos with adjustment for teleoperations.

roboticist said...

Look up some videos of Kawada Nextage, ABB Frida, Rethink Robotics Baxter etc.

Combined with some really good manipulators like DARPA ARM-H or Robotiq hand that just helped quite a few DRC teams score high, or very advanced fully anthropomorphic hands like EluMotion, DLR HIT II, TUAT/Karlsruhe etc there really is almost no limit to what can be done.

The problem with space robots is radiation, and robots really do require a lot of modern electronics, rad hardened parts and designs wont be able to obtain the performance.
So if you want to use state of the art dexterous robot on the moon, you probably want to erect a radiation shelter for it, first.

Warren Platts said...

Re: the problem of dust. Lunar dust is so abrasive primarily because it's never been weathered. But dust created in an underground hard rock mine is also unweathered, and should be of comparable abrasiveness WRT lunar dust. I agree with Baiden that it's a manageable problem.

Doug Plata said...

Several points.

We really, really need to disabuse ourselves of the idea that lunar ice harvesting is like terrestrial rock mining. Many of the breakdowns are due to the ruggedness of hammering through rock. Lunar ice harvesting might be something like driving over regolith, microwaving it, and collecting the resulting steam.

Secondly, the ore concentration on the Moon is way, way higher than terrestrial ore. So we won't have to process anywhere near as much dirt. At $5,000 to $10,000 per kg, it is like gold sitting in the surface if the Moon. Terrestrial mining analogies are not analogous.

It makes sense (to me) to first send telerobots. There would be two types: A harvester/steamer/transporter and a dexterous telerobot. The later would be used only rarely as needed for set-up and basic repairs. The harvester would be designed in such a way to make breakdowns as rare as possible and repairs as easy to accomplish by the dexterous telerobot as possible. Drop a hexnut? You can remove the tire from a modern bike without being able to drop the nut and you don't even need a wrench to do that either. So let's think smart when we design the harvesting equipment.

Given the costs and risks when sending people, in early operations, you send spare parts and redundant equipment as needed. Meanwhile, the same landers used to send redundant equipment and spare parts are getting flight experience which reduces the risk for when you later send crew on those very same landers. Crew are needed to redeem the value of any equipment that couldn't be fixed by the dexterous telerobots, to produce bulky metal parts for an expanding telerobotic workforce, and to settle / colonize.

Lunar dust is fractured glass which may be different than underground hard rock.

Interplanetary probes get 4.6 times the radiation as does electronic equipment on the ISS. In both cases, electronics can work for years. I don't think we need radiation shelters for telerobots. They just need to be rad hardened.

The best video that I have found demonstrating dexterous telerobot can be found by YouTubing: Justin telerobotics. Sorry, can't paste the link using my iPhone.



Edmund said...

Sorry to necro this, but I participated in NASA studies for robotics on the moon. The biggest problem is lunar dust. It's not just fragments of glass, but has small jagged pieces of rock and metal (incl. titanium), pieces that look like spiked balls, etc. Plus the electrostatic charge it picks up makes it hard to clean off. It's damn near the perfect stuff to gum up joints. Lunar bulldozers would be breaking down all the time.

There is also the problem that only the top layer of the regolith is loose. A short distance down it gets very, very hard to dig up, as it is densely packed due to the shaking caused by microquakes from meteoric impacts. It's denser than wet beach sand. It will be very hard to dig up. The best mining idea is probably a dragline.

Anonymous said...

Saw your argument with some guy about how to properly argue the orbital speed of the moon (at "space station" orbit). Couldn't find a way to post, but I wanted to say kudos. The gravity equation, by the way, is how the masses of the planets were calculated...

Anyway, its amusing how many people argue that robots are all that are needed. Today, after a short period on the moon, the Chinese lander died. That's why you need humans. Good point.