Author Topic: Dispelling the myth of robotic efficiency  (Read 46994 times)

Offline simonbp

  • Science Guy
  • Senior Member
  • *****
  • Posts: 7138
  • Liked: 314
  • Likes Given: 183
Dispelling the myth of robotic efficiency
« on: 03/29/2012 08:29 pm »
(Putting this in Moon, as his main comparison point is Apollo, but also applies to Mars)

http://arxiv.org/abs/1203.6250

Offline QuantumG

  • Senior Member
  • *****
  • Posts: 9238
  • Australia
  • Liked: 4477
  • Likes Given: 1108
Re: Dispelling the myth of robotic efficiency
« Reply #1 on: 03/29/2012 10:50 pm »
Great article, strange place to distribute it.

Human spaceflight is basically just LARPing now.

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #2 on: 03/30/2012 04:26 am »
Excellent paper.  Nothing strange about where it will be published - A&G is widely read.
Apologies in advance for any lack of civility - it's unintended

Offline Proponent

  • Senior Member
  • *****
  • Posts: 7277
  • Liked: 2782
  • Likes Given: 1462
Re: Dispelling the myth of robotic efficiency
« Reply #3 on: 03/31/2012 03:49 pm »
The paper often expresses efficiency in terms of science per unit mission time.  It mentions, for example, Steve Squyres' observation that the MERs do in a sol what a geologist could do in less than a minute.  Another example is Figure 3, which shows separately the cumulative number of publications per day of field work for both Apollo and for the Lunokhods.  The trouble is, the limiting resource is not mission time, but money.  By itself, efficiency expressed as scientific output per unit mission time is completely irrelevant.

The paper also inflates Apollo costs to the present with an irrelevant inflation index: the inflation rate relevant to the space sector has been quite a bit higher than indicated by any broad-economy index.  The cost of Apollo is thus underestimated.

To some extent, the paper also compares robotic martian exploration with manned lunar exploration (e.g., MERs vs. Apollo).  Well, human exploration of Mars would be quite a bit more expensive than that of the moon.

The title claims that "human space exploration will tell us more about the Solar System than will robotic exploration alone."  That may well be true, but I suspect it will be in large part because people will be willing to spend far more money on human exploration than on robotic exploration.
« Last Edit: 03/31/2012 03:50 pm by Proponent »

Offline Warren Platts

Re: Dispelling the myth of robotic efficiency
« Reply #4 on: 03/31/2012 05:38 pm »
(Putting this in Moon, as his main comparison point is Apollo, but also applies to Mars)

http://arxiv.org/abs/1203.6250

The paper makes the same point I've been making for some time now.

The paper often expresses efficiency in terms of science per unit mission time.  It mentions, for example, Steve Squyres' observation that the MERs do in a sol what a geologist could do in less than a minute.  Another example is Figure 3, which shows separately the cumulative number of publications per day of field work for both Apollo and for the Lunokhods.  The trouble is, the limiting resource is not mission time, but money.  By itself, efficiency expressed as scientific output per unit mission time is completely irrelevant.

The question is whether HSF funds should be further gutted to fund robotic science missions, e.g., the constant call for "robotic precursors" to go to the Moon before humans. If humans are more efficient at doing science than robots, then there is little point to robotic precursors.

Quote
The paper also inflates Apollo costs to the present with an irrelevant inflation index: the inflation rate relevant to the space sector has been quite a bit higher than indicated by any broad-economy index.  The cost of Apollo is thus underestimated.

Yeah, but the costs of many of the components, like electronics and computers. Also a big part of the Apollo cost was building the infrastructure at KSC and elsewhere. These costs wouldn't have to be incurred (as much) by future missions to the Moon.

Also, the paper talks about the total aggregate costs of Apollo and then divides by the number of visits. Really, we should be more interested in the incremental costs for a Lunar mission. If we can get the flight rate up enough to where we're doing 3 to 5 missions per year, the cost of a human mission to the Moon gets down to the cost of a single MSL mission.

Since Mars just isn't practical for HSF over the short to intermediate term, robots are all we are going to get for the time being, and we're going to have to be satisfied with them. The real question is the role of robotics in a major Lunar program; e.g., must we send a LPVE to the Lunar polar craters before we send a human there? LPVE was projected to cost over a billion; probably if it was ever launched it would be $2B at least by the time everything was said and done.

Well, why not? The problem is the decadal survey SMD guys don't like funding expensive flagship missions to the Moon, because they would have to sacrifice a mission to more enticing targets like Enceladus or whatever the latest moon du jour is. So to fund the mission, either extra money would have to come from Congress, or else it would have to come from the HSF budget as a "robotic precursor" mission. Similarly, there is the Spudis and Lavoie proposal to fund 16 years of robotic missions to build up a base before the first humans are sent. Me personally, we should make getting human boots on the ground a priority ASAP. Once humans are available, there's little need for robotic rovers IMHO.

Quote
To some extent, the paper also compares robotic martian exploration with manned lunar exploration (e.g., MERs vs. Apollo).  Well, human exploration of Mars would be quite a bit more expensive than that of the moon.

Not necessarily, if the cost of spaceflight can be brought down enough: e.g., going with a depot-based architecture, mostly reusable spacecraft, Lunar propellant, commercial launch, SEP's, etc.

Quote
The title claims that "human space exploration will tell us more about the Solar System than will robotic exploration alone."  That may well be true, but I suspect it will be in large part because people will be willing to spend far more money on human exploration than on robotic exploration.

This point was addressed somewhat in the paper by pointing out there was a lot more to the Apollo missions and HSF in general than planetary science. The HSF program is going to carry on no matter what, whether it lands on the Moon or not, or merely engages in more literal navel gazing at an EML1 space station or something. If we're interested in planetary science, we should encourage the decisionmakers to send people where they can do the most scientific good. At this point, that's the Moon; NEA missions are going to be few and far between; Mars is a bridge too far. Yes, that entails building a lander.
"When once you have tasted flight, you will forever walk the earth with your eyes turned skyward, for there you have been, and there you will always long to return."--Leonardo Da Vinci

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #5 on: 03/31/2012 11:12 pm »
The paper often expresses efficiency in terms of science per unit mission time.  It mentions, for example, Steve Squyres' observation that the MERs do in a sol what a geologist could do in less than a minute.  Another example is Figure 3, which shows separately the cumulative number of publications per day of field work for both Apollo and for the Lunokhods.  The trouble is, the limiting resource is not mission time, but money.  By itself, efficiency expressed as scientific output per unit mission time is completely irrelevant.

What efficiency indices would you suggest?

Quote
The paper also inflates Apollo costs to the present with an irrelevant inflation index: the inflation rate relevant to the space sector has been quite a bit higher than indicated by any broad-economy index.  The cost of Apollo is thus underestimated.

All missions examined (which include unmanned missions) use the same index, so whether or not this is the best index is a secondary issue. What matters is the proportional costs.

Quote
To some extent, the paper also compares robotic martian exploration with manned lunar exploration (e.g., MERs vs. Apollo).  Well, human exploration of Mars would be quite a bit more expensive than that of the moon.

A point the author makes.  However the study primarily compares the productivity of unmanned lunar missions with manned lunar missions and then extrapolates these to Mars.


Apologies in advance for any lack of civility - it's unintended

Offline IRobot

  • Full Member
  • ****
  • Posts: 1312
  • Portugal & Germany
  • Liked: 310
  • Likes Given: 272
Re: Dispelling the myth of robotic efficiency
« Reply #6 on: 04/01/2012 12:54 am »
Artificial intelligence, image recognition and robotics have had a tremendous development in the past 10 years. If man is not in the way, it will be in a couple of years.
I sure hope manned exploration proceeds, cause there is nothing exciting with robots. But soon there won't be a thing that can't be done by a robot instead of a man.

Offline QuantumG

  • Senior Member
  • *****
  • Posts: 9238
  • Australia
  • Liked: 4477
  • Likes Given: 1108
Re: Dispelling the myth of robotic efficiency
« Reply #7 on: 04/01/2012 01:12 am »
That presupposes that space is on the forefront of technology.. it isn't. Although it is quite possible to send a fully automated robot to Mars that can do good science, that isn't what is done and won't be, so long as the motivation for sending these probes is the careers of scientific researchers here on Earth.

Human spaceflight is basically just LARPing now.

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #8 on: 04/01/2012 02:37 am »
Artificial intelligence, image recognition and robotics have had a tremendous development in the past 10 years. If man is not in the way, it will be in a couple of years.
I sure hope manned exploration proceeds, cause there is nothing exciting with robots. But soon there won't be a thing that can't be done by a robot instead of a man.

People have been predicting this for fifty years and we are relatically nowhere close to that goal, despite staggering increases in computing power.
Apologies in advance for any lack of civility - it's unintended

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #9 on: 04/01/2012 02:40 am »
That presupposes that space is on the forefront of technology.. it isn't. Although it is quite possible to send a fully automated robot to Mars that can do good science, that isn't what is done and won't be, so long as the motivation for sending these probes is the careers of scientific researchers here on Earth.


The motivation of most scientists in sending these probes is curosity not their careers.  If they were interested in their careers they would chose a much more rewarding profession than space exploration, or even science.

There is no evidence what so ever that "fully automated robot" could equal a what an astronaut could achieve and a lot of evidence to the contrary, as Crawford's paper amply demonstrates. 

EDIT Additional text
« Last Edit: 04/01/2012 05:12 am by Dalhousie »
Apologies in advance for any lack of civility - it's unintended

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5152
  • Liked: 1002
  • Likes Given: 342
Re: Dispelling the myth of robotic efficiency
« Reply #10 on: 04/01/2012 03:44 am »
Good luck putting a human geologist on mars who would observe local weather patterns for 8 years without any resupplies.

( there is an obvious point to this post, i'd rather not state it )
Orion - the first and only manned not-too-deep-space craft

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #11 on: 04/01/2012 05:08 am »
Good luck putting a human geologist on mars who would observe local weather patterns for 8 years without any resupplies.

( there is an obvious point to this post, i'd rather not state it )

If you just want basic met data, use a an automated station.  On the other hand the geologist, doing stuff only a greologist can do, if  properly trained might well pick up meterological phenomena that the robot could not, because they occur outside the observation parameters of the system.

There are some things that automation is very good at, others that ity is not.


But Crawford's paper shows very clearly is that that people excel, by orders of magntitude, atexploration, and that the fruits of human presence continue to pay dividends decades down the track.
Apologies in advance for any lack of civility - it's unintended

Offline Proponent

  • Senior Member
  • *****
  • Posts: 7277
  • Liked: 2782
  • Likes Given: 1462
Re: Dispelling the myth of robotic efficiency
« Reply #12 on: 04/01/2012 12:35 pm »
The paper often expresses efficiency in terms of science per unit mission time.  It mentions, for example, Steve Squyres' observation that the MERs do in a sol what a geologist could do in less than a minute.  Another example is Figure 3, which shows separately the cumulative number of publications per day of field work for both Apollo and for the Lunokhods.  The trouble is, the limiting resource is not mission time, but money.  By itself, efficiency expressed as scientific output per unit mission time is completely irrelevant.

What efficiency indices would you suggest?

I would suggest that it's up to the Crawford to define what kind of efficiency he's talking about, since he states that robotic probes are less efficient.  He doesn't do that but rather throws around various ideas of efficiency without examining whether they're relevant.  I suggest that output per dollar spent is the relevant criterion, since money is almost always the limiting factor.  Crawford doesn't seem much interested in this measure.

Quote
Quote
The paper also inflates Apollo costs to the present with an irrelevant inflation index: the inflation rate relevant to the space sector has been quite a bit higher than indicated by any broad-economy index.  The cost of Apollo is thus underestimated.

All missions examined (which include unmanned missions) use the same index, so whether or not this is the best index is a secondary issue. What matters is the proportional costs.

The point of inflating costs to the present is to allow apples-to-apples comparisons of costs incurred at different times.  Granted, a comparison of the cost of Surveyor to that of Apollo is relatively insensitive to the inflation index used, since both programs occurred at about the same time.  If, however, we're comparing Apollo with LRO, then the inflation index used makes a difference.  Between 1966 and 2010, the CPI (used by Crawford) rose by a factor of about 6.7.  The more appropriate NASA New-Start Index rose by a factor of 8.9.  (Now that I've calculated those ratios, I admit the difference isn't as large as I expected it to be, somehat blunting my criticism.)

Again, my point is that Crawford never adopts a definition of efficiency and presents various measures that might be regarded as efficiencies but aren't really relevant.
« Last Edit: 04/01/2012 12:39 pm by Proponent »

Offline Proponent

  • Senior Member
  • *****
  • Posts: 7277
  • Liked: 2782
  • Likes Given: 1462
Re: Dispelling the myth of robotic efficiency
« Reply #13 on: 04/01/2012 12:42 pm »
Please recall, everybody, that one of the most astounding discoveries about the moon, namely the presence of polar volatiles, has come entirely from robotic exploration.

Offline Warren Platts

Re: Dispelling the myth of robotic efficiency
« Reply #14 on: 04/01/2012 04:56 pm »
Please recall, everybody, that one of the most astounding discoveries about the Moon, namely the presence of polar volatiles, has come entirely from robotic exploration.

Actually, the polar volatiles were predicted by theory since at least the 1960's. Also, you're conflating orbiters with rovers. No one's arguing that orbiters are less efficient than humans in orbit.

But eventually, you reach a point of diminishing returns with orbiters. There can be no question that a LOT was learned about Apollo, and indeed papers are still being published, most notably the discovery the volcanic glass beads actually contained water levels comparable to the Earth's mantle, indicating that the Moon probably has significant endogenous water supplies, such that outgassing my still be an ongoing process.

It's surface operations where the claim is made that humans are more efficient on a $$$/scientific results basis.

Granted, there was the LCROSS results, which provided some big surprises, especially all the metals (although the mercury results were predicted by George Reed in the 1990's.)  And the LCROSS experiment was a veritable bargain (and I think we should try and replicate the experiment again).  However, there are still huge uncertainties surrounding the results: e.g., the authors issued an erratum after their article was published that reduced their reported concentrations by a factor of 5, but even with this qualification the concentrations are still way too high most likely (Paul Spudis, personal communication).

This is a major problem for just about all robotic lander (or crasher) missions: they deliver tantalizing results that aren't definitive. We'll never know what Cabeus Crater is really like until we send in some humans.

I will further grant, however, that the efficiency of human exploration depends a lot on the chosen architecture. E.g., the latest DRM that Chris Bergin reported on is going to be hugely expensive: e.g., leaving out the development costs of a lander (which could easily run $10B), one mission every 4 years is going to have an incremental cost on the order of 12 to 24 $B/ea.; at such cost levels, you can question whether human exploration is the more efficient. In which case, we're probably better off going with the Spudis and Lavoie approach that attemps to use robots to leverage Lunar ISRU propellants to lower the cost of human missions...
"When once you have tasted flight, you will forever walk the earth with your eyes turned skyward, for there you have been, and there you will always long to return."--Leonardo Da Vinci

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #15 on: 04/02/2012 04:02 am »
Please recall, everybody, that one of the most astounding discoveries about the moon, namely the presence of polar volatiles, has come entirely from robotic exploration.

Note:

1) Important yes, but not astounding - it was predicted decades ago from basic physics.  It would have been astounding if no polar ice were found.

2) In principle it could have been discovred by earth-based radar (as indeed were the polar ice deposits of Mercury).  Should this be used as an argument against orbiters.

3) The efficiencies that Crwaford discusses pertained to surface exploration, not orbiters.  I don't think anyone claims that remote sensing from orbit would be better undertaken by crewed missions

« Last Edit: 04/02/2012 04:02 am by Dalhousie »
Apologies in advance for any lack of civility - it's unintended

Offline savuporo

  • Senior Member
  • *****
  • Posts: 5152
  • Liked: 1002
  • Likes Given: 342
Re: Dispelling the myth of robotic efficiency
« Reply #16 on: 04/02/2012 04:08 am »
If you just want basic met data, use a an automated station.
You mean, a robot ? : )

Quote
There are some things that automation is very good at, others that ity is not.

Thanks for stating the obvious.
Orion - the first and only manned not-too-deep-space craft

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #17 on: 04/02/2012 05:07 am »

I would suggest that it's up to the Crawford to define what kind of efficiency he's talking about, since he states that robotic probes are less efficient.  He doesn't do that but rather throws around various ideas of efficiency without examining whether they're relevant.  I suggest that output per dollar spent is the relevant criterion, since money is almost always the limiting factor.  Crawford doesn't seem much interested in this measure.

Crawford lists a whole range of criteria by which efficiency can be measured, 11 by my count. 

Page 1

1 On the spot decision making and flexibility (extensively documented in other literature)

2 Mobility (documented on pages 2-3)

3 Number of sites visited, samples collected, and mass samples returns (documented on pages 3-4)

Page 2

4 Ability to perform complex activities and deploy complex equipment (geophysical arrays, deep drills)

5 Ability to deploy and expand large scale infrastructure (typified by Hubble)

6 Performance of key tasks needed for planetary exploration (documented in Figure 1 on page 10)

7 Number of sites visited per mission

Page 4

8 Scientific productivity measured by total cumulative number of refereed publications (documented in figure 2 on page 10)

9 Scientific productivity measured by cumulative number per day of refereed publications (documented in figure 3 on page 11)

9 Scientific productivity measured by time scale over which mission generates research after its conclusion (documented in figure 2 on page 10)

Page 5

10 Performance compared on basis of cost per mission

11 Performance based on science cost per site

Crawford certainly does not just “throws around various ideas of efficiency without examining whether they're relevant.”

Quote
I suggest that output per dollar spent is the relevant criterion, since money is almost always the limiting factor.  Crawford doesn't seem much interested in this measure.

You could come up with a huge number of indices, you can’t expect a short paper in A&G to cover every possibility.  Bute let’s do what you suggest.

The Surveyor program is said by the font of all knowledge (Wikipedia) to have cost $469 million at the time.  From Figure 2 (and it’s rather small) it would appear that there were about 50 refereed Surveyor papers, that’s 7.38 million a paper. Apollo cost $25 billion in the same time frame, the ~2800 papers in Figure 2 that comes to 8.93 million a paper.  That makes Apollo about 20% more expensive, per paper.  On the hand the science done by Apollo was also much more sophisticated and continues to yield dividends today, Surveyor’s science dried up in about 15 years (again from Figure 2).

If you accept Crawford’s argument that the bulk of Apollo funding was geopolitical and only the science component should be factored, then the entire science of Apollo ($387 million) was less than the Surveyor program.  This would make return cost per Apollo paper $138 thousand.

Another index you could use would be cost per site.  Surveyor visited five sites, Apollo more than 2000.  That’s more than $93 million a site against less than $12.5 million a site.  Again, we have high resolution imagery and samples from every Apollo site yielding data vastly superior to that obtained by Surveyor.

Quote
The paper also inflates Apollo costs to the present with an irrelevant inflation index: the inflation rate relevant to the space sector has been quite a bit higher than indicated by any broad-economy index.  The cost of Apollo is thus underestimated.

The point of inflating costs to the present is to allow apples-to-apples comparisons of costs incurred at different times.  Granted, a comparison of the cost of Surveyor to that of Apollo is relatively insensitive to the inflation index used, since both programs occurred at about the same time.  If, however, we're comparing Apollo with LRO, then the inflation index used makes a difference.  Between 1966 and 2010, the CPI (used by Crawford) rose by a factor of about 6.7.  The more appropriate NASA New-Start Index rose by a factor of 8.9.  (Now that I've calculated those ratios, I admit the difference isn't as large as I expected it to be, somewhat blunting my criticism.)

Thanks for this number, where did you get it from?

Interestingly it makes the cost of the Surveyor program $4.17 billion, which is an thought provoking comparison, making each of the seven Surveyor missions costing about the same as Phoenix lander or a MER. 

Using this for Apollo prices it in today’s terms about $223 billion, about 89 times that of MSL, 14.8 times per landing site.  We will have to wait and see whether the productivity of MSL will be one 15th of any given Apollo mission.

Again Crawford would argue that possibly only the science component should be included, that comes to $3.44 billion in today’s terms, 1.4 times MSL.

Quote
Again, my point is that Crawford never adopts a definition of efficiency and presents various measures that might be regarded as efficiencies but aren't really relevant.

He proposes 11 different quantitative measures, three of which have been used by previous workers, which he cites.
Apologies in advance for any lack of civility - it's unintended

Offline Dalhousie

  • Senior Member
  • *****
  • Posts: 2766
  • Liked: 780
  • Likes Given: 1131
Re: Dispelling the myth of robotic efficiency
« Reply #18 on: 04/02/2012 05:14 am »
If you just want basic met data, use a an automated station.
You mean, a robot ? : )

If you like.  But the problem with the word "robot" is that it is highly loaded and arouses unrealistic expectations.

automatic weather stations and geophyscial observatories are commonplace on Earth.  They are not generally called robots.

Quote
There are some things that automation is very good at, others that it is not.

Quote
Thanks for stating the obvious.

Sometimes the obvious needs to be stated. 

I have plenty more where that came from, all from robotics researchers.

Here is another one.  What is easy for robots is hard for people and what is easy for people is hard for robots.
Apologies in advance for any lack of civility - it's unintended

Offline A_M_Swallow

  • Elite Veteran
  • Senior Member
  • *****
  • Posts: 8906
  • South coast of England
  • Liked: 500
  • Likes Given: 223
Re: Dispelling the myth of robotic efficiency
« Reply #19 on: 04/02/2012 09:59 am »
If you just want basic met data, use a an automated station.
You mean, a robot ? : )

Robots are assumed to move.  An automated station on the lunar surface is likely to stay in one place.

Tags:
 

Advertisement NovaTech
Advertisement Northrop Grumman
Advertisement
Advertisement Margaritaville Beach Resort South Padre Island
Advertisement Brady Kenniston
Advertisement NextSpaceflight
Advertisement Nathan Barker Photography
0