Dispelling the myth of robotic efficiency

There is a widely held view in the astronomical community that unmanned robotic space vehicles are, and always will be, more efficient explorers of planetary surfaces than astronauts (e.g. Coates 2001, Clements 2009, Rees 2011). Partly this comes from a common assumption that robotic exploration is cheaper than human exploration (although this isn't necessarily true if like is compared with like) and partly from the expectation that developments in technology will relentlessly increase the capability and reduce the size and cost of robotic missions to the point that human exploration will not be able to compete. I argue below that the experience of human exploration during the Apollo missions, more recent field analogue studies and trends in robotic space exploration all point to exactly the opposite conclusion. As demonstrated by the Apollo missions 40 years ago — and leaving the question of cost to a separate examination below — human space exploration has a number of advantages over robotic operations on planetary surfaces. These have been discussed in detail elsewhere (e.g. Spudis 2001, Crawford 2001, 2010, Garvin 2004, Cockell 2004, Snook et al. 2007) and were endorsed by the independent Commission on the Scientific Case for Human Space Exploration commissioned by the RAS in 2005 (Close et al. 2005, hereinafter “the RAS Report”). These advantages can be summarized as follows: On-the-spot decision making and flexibility, with increased opportunities for serendipitous discoveries. Greatly enhanced mobility and attendant opportunities for geological exploration and instrument deployment. Compare the 35.7 km traversed in three days by the Apollo 17 astronauts in December 1972 with the almost identical distance (34.4 km) traversed by the Mars Exploration Rover Opportunity in eight years. Greatly increased efficiency in sample collection and sample return capacity. Compare the 382 kg of samples returned by Apollo with the 0.32 kg from the Russian robotic sample return missions Lunas 16, 20 and 24, and the zero kg returned so far by any robotic mission to Mars. Increased potential for large-scale exploratory activities (e.g. drilling) and the deployment and maintenance of complex equipment. The development of a space-based infrastructure to support space-based astronomy and other scientific applications (e.g. the construction and maintenance of large space telescopes). The increasing size of Mars rovers, from Pathfinder (front left), a Mars Exploration Rover (left), to Mars Science Laboratory (right). This increase in size (and cost), contrary to predictions that improved technology will result in smaller and cheaper robots, is mandated by the nature of the martian surface and complexity of exploration objectives. Human missions would be even larger and more expensive, but, crucially, much more capable. (NASA/JPL-Caltech) With the exception of the final point, for which the best demonstration is provided by the five successful space shuttle servicing missions to the Hubble Space Telescope (e.g. NRC 2005), demonstration of the benefits of human spaceflight for planetary exploration must be sought in a comparison of the relative efficiencies of the Apollo missions and robotic missions to the Moon and Mars, supported where appropriate with terrestrial analogue studies. The relative efficiency of human over robotic exploration of planetary surfaces is well recognized by scientists directly involved with the latter. For example, regarding the exploration of Mars, the RAS Report (paragraph 70) noted: “The expert evidence we have heard strongly suggests that the use of autonomous robots alone will very significantly limit what can be learned about our nearest potentially habitable planet.” Putting it more bluntly, Steve Squyres, the PI for the Mars Exploration Rovers (MERs) Spirit and Opportunity, has written: “[t]he unfortunate truth is that most things our rovers can do in a perfect sol [i.e. a martian day] a human explorer could do in less than a minute” (Squyres 2005 p234–5). This is of course only a qualitative assessment, albeit by someone well placed to make an informed judgement. Nevertheless, at face value it implies a human/robot efficiency ratio of about 1500, which is far larger than the likely ratio of cost between a human mission to Mars and the cost of the MERs (see below). Even this, however, does not fairly compare human exploration efficiency with robotic exploration. This is because much of the scientific benefit of human missions will consist of samples returned, drill cores drilled and geophysical instruments deployed, all of which were demonstrated by Apollo on the Moon, but none of which have been achieved by the MERs nor will be achieved by the more capable (and vastly more expensive) Mars Science Laboratory (MSL) that is due to land on Mars in 2012. More objective estimates of the relative efficiency of robots and humans as field geologists have been given by Garvin et al. (2004)) and Snook et al. (2007)). Garvin (2004) summarized the results of a NASA survey of several dozen planetary scientists and engineers on the relative efficiency of human and robotic capabilities in 18 different skill sets relevant to planetary exploration. The results are summarized in table 1, and show a clear balance in favour of human capabilities (with the implicit recognition that the most efficient exploration strategies of all will be those consisting of human—robotic partnerships where each complements the other). Humans vs robots in space Tabular summary of exploration skills after Garvin (2004). The fourth column indicates relative advantage of humans or robots on a sliding scale (green: the advantage lies with humans; red: with robots; white: equal). In most cases humans have a clear advantage. Since 2004 the extreme endurance of the MERs has moved this entry more in favour of robots, but they remain slow and inflexible. (Courtesy of Jim Garvin/NASA/Springer) Humans vs robots in space Tabular summary of exploration skills after Garvin (2004). The fourth column indicates relative advantage of humans or robots on a sliding scale (green: the advantage lies with humans; red: with robots; white: equal). In most cases humans have a clear advantage. Since 2004 the extreme endurance of the MERs has moved this entry more in favour of robots, but they remain slow and inflexible. (Courtesy of Jim Garvin/NASA/Springer) This conclusion is corroborated by direct field comparisons of human and robotic exploration at planetary analogue sites on Earth. Snook et al. (2007)) reported the results of one such study, conducted at the Haughton impact crater in the Canadian Arctic, where the efficiency of a human explorer (suitably encumbered in a spacesuit) was compared with that of a tele-operated rover (controlled from NASA Ames Research Centre in California) in the performance of a range of exploration tasks. The rover was more sophisticated than those employed in present-day space missions, and included simulation of artificial intelligence capabilities that are only likely to be incorporated in actual space missions from 2015 at the earliest. Nevertheless, the space-suited “astronaut” was found to be much more efficient in performing exploration tasks than the rover, and Snook et al. (2007 p438) concluded that “humans could be one to two orders of magnitude more productive per unit time in exploration than future terrestrially controlled robots”. Although this estimate is an order of magnitude lower than Squyres’ off-the-cuff estimate of 1500 given above, this is mainly because the comparison was conducted between human and tele-robotic exploration, rather than between humans and supervised quasi-autonomous robotic exploration such as carried out by the MERs and MSL. Tele-robotic exploration is known to be more efficient than autonomous robotic operation, precisely because real-time human interaction is involved, but it cannot be employed effectively on planetary surfaces more distant than the Moon because of the inevitable communications delay (Lester and Thronson 2011). Garvin (2004, see his figure 2) has compared the efficiencies of robotic, tele-robotic, and human exploration, from which it is clear that if humans are “one to two orders of magnitude more efficient” than tele-robots then they will be even more efficient when compared with robotic vehicles such as the MERs or MSL, bringing the two estimates into better agreement. Cumulative number of refereed publications in the ADS database for the Apollo, Luna 16/20/24, Lunokhod, Surveyor, and Mars Exploration Rover missions. See “Bibliometric search details” on page 2.26 for more information. Moreover, while comparisons based on the relative time taken to perform certain tasks do indeed show humans to be more efficient than robots, they nevertheless grossly underestimate the added scientific value of having humans on planetary surfaces. This is because astronauts have to come back to Earth, and can therefore bring large quantities of intelligently collected samples back with them. Robotic explorers, on the other hand, generally do not return — one reason why they are cheaper — so nothing can come back with them. Even if robotic sample return missions are implemented, neither the quantity nor the diversity of these samples will be as high as would be achievable in the context of a human mission — again compare the 382 kg of samples (collected from more than 2000 discrete locations) returned by Apollo, with the 0.32 kg (collected from three locations) brought back by the Luna sample return missions. The Apollo sample haul might also be compared with the =0.5 kg generally considered in the context of future robotic Mars sample return missions (e.g. ISAG 2011). Note that this comparison is not intended in any way to downplay the scientific importance of robotic Mars sample return, which will in any case be essential before human missions can responsibly be sent to Mars, but merely to point out the step change in sample availability (both in quantity and diversity) that may be expected when and if human missions are sent to the planet. If, as argued above, human exploration of planetary surfaces is really scientifically so much more efficient than robotic exploration, then we would expect to see this reflected in the scientific literature. As the only human exploration missions to date are the Apollo missions, any such bibliometric comparison must be between publications based on Apollo data and those based on various robotic missions to the Moon and Mars. With this in mind, figure 2 shows the cumulative number of refereed publications recorded in the SAO/NASA Astrophysics Data System (ADS) resulting from the six successful Apollo landings, the three Luna sample return missions (Lunas 16, 20, 24), the two tele-operated Lunokhod rovers (Lunas 17, 21), the five successful Surveyor lunar soft landers, and the two MERs (Spirit and Opportunity) on Mars. Those interested will find more details of the search parameters in the box on page 2.26. Several things are immediately apparent. Most obvious is the sheer volume of Apollo's scientific legacy compared to the other missions illustrated. This alone goes a long way to vindicate the points made above about human versus robotic efficiency. The second point to note is that the next most productive set of missions are the lunar sample return missions Lunas 16, 20 and 24, which highlights the importance of sample return. Indeed, a large part of the reason why Apollo has resulted in many more publications than the Luna missions is the much larger quantity and diversity of the samples — which will always be greater in the context of human missions. The third point to note is that, despite being based on data obtained and samples collected more than 40 years ago, and unlike the Luna, Lunokhod, or Surveyor publications, which have clearly levelled off, the Apollo publication rate is still rising. Indeed, it is actually rising as fast as, or faster than, the publications rate derived from the Mars Exploration Rovers, despite the fact that data derived from the latter are much more recent. No matter how far one extrapolates into the future, it is clear that the volume of scientific activity generated by the MERs, or other robotic exploration missions, will never approach that due to Apollo. However, to my mind, the most staggering thing about figure 2 is that this enormous scientific legacy is based on a total of only 12.5 days total contact time with the lunar surface. Note that this is the total cumulative time the Apollo astronauts were on the Moon, including down-time in the Lunar Module, not the cumulative EVA time, which was just 3.4 days (Orloff and Harland 2006). This may be compared with a total of 436 active days on the surface for the Lunokhods (Wilson 1987) and 5162 days for the Mars Exploration Rovers (to the end of 2011, allowing for the fact that contact was lost with Spirit on 22 March 2010). This comparison is illustrated in figure 3, which shows the cumulative number of publications divided by days of fieldwork on the surface (adopting 12.5 days for Apollo to allow for a fair comparison with the rovers). This is the same as dividing the cumulative number of publications by the number of sites studied up to a given date and the average days of fieldwork per site. Dips in the cumulative curves occur when a new mission arrives, instantaneously increasing the days of fieldwork before this has fed through to increased publications. Note the logarithmic scale — by this metric, Apollo was over three orders of magnitude more efficient in producing scientific papers per day of fieldwork than are the MERs. This is essentially the same as Squyres’ (2005)) intuitive estimate given above, and is consistent with the more quantitative analogue fieldwork tests reported by Snook et al. (2007)). Refereed publications per day exploring the surface of the Moon or Mars. This is the same as dividing the cumulative number of publications by the number of sites visited by a given date and the average time spent there. The plot covers those missions plotted in figure 2 which had, or have, surface mobility, and which can therefore be considered as having performed “fieldwork”. Although it is generally taken for granted that human exploration is more expensive than robotic exploration, and this is certainly true if the aggregate costs are the only ones considered, the situation is not as clear cut as it is sometimes made out to be. For one thing, the ratio of costs between human and robotic missions, while large, may nevertheless be smaller than the corresponding ratio in scientific productivity. The Apollo missions are instructive in this respect. Wilhelms (1993) and Beattie (2001) estimated a total cost of Apollo as $25bn “in 1960s money”. This is rather more than the Congressional appropriations for Apollo ($19.4bn from 1961 to 1973; tabulated by Orloff and Harland 2006). Taking the higher estimate (to be conservative) and taking “1960s” to be 1966 when Apollo expenditure peaked, this corresponds to about $175bn today (where I have made use of the US Bureau of Statistics Inflation Calculator at ). It is interesting to compare this with the cost of a modern state-of-the-art robotic mission such as Mars Science Laboratory. MSL, which at this writing is en route to Mars, has cost an estimated $2.5bn (Leone 2011). Thus, in real terms, Apollo cost 70 times as much as MSL. However, Apollo visited six sites, whereas MSL will visit one. In terms of cost-per-site Apollo was only 12 times dearer than MSL, yet each Apollo mission was vastly more capable. It is true that this comparison only strictly holds in the context of lunar exploration, where we can compare Apollo with a hypothetical future MSL-like lunar rover. In the context of Mars exploration, human missions seem likely to be more expensive than Apollo in real terms, although not necessarily by a large factor. The estimated total costs of some human Mars mission architectures are comparable to that of Apollo, or even lower (Turner 2004). The main point is that human missions like Apollo are between two and three orders of magnitude more efficient in performing exploration tasks than robotic missions, while being only one to two orders of magnitude more expensive. In addition, human missions can accomplish scientific objectives that are unlikely to be achieved robotically at all (deep drilling and properly representative sample collection and return are obvious examples, as well as the increased opportunities for serendipitous discoveries). Looked at this way, human space exploration seem so expensive after is a more sophisticated and productive way to view the relative costs of human and robotic The fact is that while robotic planetary missions are and essentially costs are therefore by scientific human spaceflight is not or even the of human spaceflight to be development and and in Thus, can be a of human missions and for by other Apollo again an it was for rather than scientific and to order the US expenditure of $25bn would have any was performed or to the of a number of scientists scientific objectives and capabilities were incorporated into Apollo and resulted in the scientific legacy that is still being including in Apollo not come It is interesting to compare this strictly with the cost of robotic missions that and are, less capable than Apollo. Beattie (2001) has studied the cost of including in Apollo the Apollo Lunar to the construction of the Lunar Laboratory in to the of lunar sample and geological and at a figure of in 1972 higher than other Beattie does not the cost of the Lunar but this be added as the was included in the three Apollo missions mainly to geological exploration. we at a total scientific cost in Apollo of in 1972 This corresponds to about or of the total Apollo It is instructive to compare this with the costs of some and robotic missions. It is sometimes to estimates for but a search of various for Mars Pathfinder on Mars in for the two Mars Exploration Rovers in 2004 and only the days of as above, $2.5bn for Mars Science Laboratory. the US Bureau of Statistics Inflation these to and $2.5bn in For to a NASA and the estimated cost of the Mars mission for is about These mission costs are compared in figure from which two points costs of robotic Mars rover missions in As two rovers including in to other expensive For how much will this be scientific Note that this is not an increase costs are we could to rovers as and as but not to to scientific The cost of the robotic exploration of Mars has not been as technology but has been increasing There is a reason for planetary surfaces are large, not at all to exploration with rovers matter how much is into them. There isn't much point in having a rover the size of a if it can only a day and time it a the size of a is such a likely to much in the way of As a rovers have and more expensive (and more with time This is exactly opposite to the by some (e.g. Rees but it is one which will if we in to planetary surfaces with In real terms, the cost of Mars Science Laboratory the cost of on Apollo in In as Apollo visited six different sites on the Moon, the cost of per in is actually less than the cost of any of these robotic Mars missions as above, the scientific efficiency was The if at some future date a of human missions return to the Moon are sent on to Mars, and if these are they will for a complex range of scientists will more for our on than we will by on autonomous robotic vehicles which in any is a way In the space to the Exploration which the for a human exploration that could with just such an of the of the has been the development of a Exploration which to human missions to the Moon, Mars. The for the are, and a range of and of in in to Science would be a of in a human exploration such as by the Exploration this will result in new including to regarding the and of and the and of in the that will not be obtained as and in many cases not obtained at by on robotic exploration The bibliometric in 2 and were obtained from the SAO/NASA ADS The were to refereed publications. It is likely that these underestimate the total number of publications resulting from these missions, for Apollo, as some sample has been in geological and not included in the ADS through the of the of the Lunar and the ADS be up to about when the Lunar was over to the 2011). I have made use of the ADS because of of and because it is a bibliometric database in the astronomical The Apollo publications were by for papers the in the or was to papers to the Apollo These were then to that they do indeed to the Apollo missions. the Surveyor papers were by for (where this time was included to to to the Surveyor The Luna and Lunokhod publications were by for these missions by the publications were by for Exploration this set of search was to the large number of papers conducted in such and such a or which for various things to Mars was made the Science publication to that these papers were by the ADS search (although the papers are only a of the total number of papers making to the MERs included in figure this that a large number of papers have not been in the ADS I and as and the for on the bibliometric This Astrophysics Data