Advanced Automation for Space Missions/Chapter 5.6

5.6 Realization

John von Neumann, and a large number of other researchers in theoretical computer science following him, have shown that there are numerous alternative strategies by which a machine system can duplicate itself. There is a large repertoire of theoretical computer science results showing how machine systems may simulate, construct, inspect, and repair machine systems including, to some extent, themselves. This repertoire may be useful in the design of actual replicating machine systems.

The basic concept of physical machines capable of useful self-replication is credible both from a theoretical and a practical engineering standpoint. It is reasonable to begin designing replicating systems based on current knowledge and state-of-the-art technology, though final design resolution will require significant additional research. Complete systems closure is achievable in principle, though partial closure may be more feasible from an economic and pragmatic engineering standpoint in the near term. It also appears feasible to begin immediate work on the development of a simple demonstration SRS on a laboratory scale, with phased steps to more sophisticated levels as the technology is proven and matures.

Self-replicating systems appear potentially useful in an economic or commercial sense. The main advantage in using SRS over other methods of space exploration and industrialization is that a very large capability for performing any desired task can be rapidly achieved at almost any remote location, starting with a relatively small investment of time, money, energy, and mass in the original "seed" mechanism. SRS will have many applications on Earth, in near-Earth and lunar space, throughout the Solar System, and in the interstellar realm, for future exploration and utilization, suggesting a number of significant social, cultural and economic impacts on American and human society.

In this section the Replicating Systems Concepts Team sets forth in some detail how NASA may take action at once toward the achievement of the ultimate goal of establishing a replicating manufacturing facility. A suggested statement of work (SOW) and a list of institutions which might undertake the tasks outlined in the work statement are included.

5.6.1 Prologue to Realization

The space program of the United States is at a critical point in its evolution. The easy missions, for the most part, have been accomplished. These have been limited to what could be done within the lift capacity of one or two launch vehicles. The capabilities of the payloads which have been delivered to space have been limited by (1) the rudimentary nature of payload automation (either preprogrammed or teleoperated), (2) the high penalty for life support systems and of man-rating manned payloads, and (3) the high cost of the Earth-based mission operations.

The industry of the U.S. is also at a critical juncture in its evolution. If it is to compete adequately in the world marketplace, significant increases in productivity are required. Present production methods have reached a level of maturity such that sufficiently large gains in productivity through further refinement of present-day technologies are unlikely to be realized. The only known solution is massive automation such as is now being applied in other industrialized countries, notably Japan and Germany.

Massive automation would dramatically increase the capabilities and effectiveness of the space program. Use of the emerging techniques of machine intelligence would make it possible to perform missions which previously would have required men in situ, thus prohibitively expensive. Highly automated programmable manufacturing by robots would permit the economical production of small numbers of spacecraft for exploratory missions. Missions which require the manipulation of large amounts of mass off-Earth (e.g., lunar/orbital bases or solar power satellites) are especially amenable to massive automation. These missions can be accomplished by employing large numbers of cheap freight rockets mass-produced by robots in automated factories and launched by robots at automated launch facilities (Cliff, Summer Study Document. 1980). These missions might also be accomplished by extraterrestrial automated manufacturing of the required hardware. In any case, the key is massive automation.

One of the most significant characteristics of massive automation is the possible regenerative or "bootstrapping" effect. Using robots to make robots will decrease costs dramatically, thus expanding the economically viable uses of robots. This in turn increases demand, leading to yet further automation, which leads to lower-cost robots, and so on. The end result is "superautomation" (Albus, 1976). A similar effect has already been noted in the computer industry where dramatic increases in performance/price have continued unabated over three decades. The use of robots to help manufacture robots, analogous to the use of computers to help make computers, should produce a similar effect. Extensive innovation should continue unabated for quite some time in such a young field.

Work is now in progress in Computer-Aided Design and Manufacture (CAD/CAM) in the United States. A partial bibliography of recent work in this area and a list of manufacturers, equipment directory, and supplier addresses have been published (Gettleman, 1979; "Numeric Control Equipment," 1980). Several bills designed to promote automation are presently before the U.S. Congress. The Department of Commerce is beginning a program to promote industrial automation in this country. The National Science Foundation also is funding work in automation. The Department of Defense has initiated a large effort in Integrated Computer-Aided Manufacturing (ICAM) (Business Week, 1980). ICAM combines both CAD and CAM (see sec. 5.4.1).

Within NASA, related work is in progress or is proposed at several Program Centers. An exhaustive search of such activities has not been possible in the limited time available, but several programs are especially noteworthy. The Jet Propulsion Laboratory has an active Advanced Development Laboratory (Bejczy, 1980). The Goddard Space Flight Center (GSFC) has proposed an effort to adapt existing CAD/CAM facilities at the Center to the control of robot manipulators for complete assembly (Purves, personal communication, 1980). Self-replicating systems have been studied at Marshall Space Flight Center (von Tiesenhausen and Darbro, 1980).

NASA unique benefits and requirements. NASA is in a unique position to benefit from massive automation - particularly self-replicating systems. The minimum possible size for a totally autonomous SRS is not presently known. However, feasibility studies performed to date (Freitas, 1980a; von Tiesenhausen and Darbro, 1980) have described systems which were quite large. Although autonomous self-replicating systems have been proposed for terrestrial use (Moore, 1956), sociocultural and ecological considerations seem to make them less practical, possibly even undesirable, on the Earth itself. This planet already supports several very large symbiotic man-machine replicating systems - the industrial societies.

In contrast to the terrestrial case, autonomous or symbiotic SRS are ideally suited to space applications. In space there is room for such systems to multiply and grow. In fact the exponentially expanding, self-replicating factory is the most promising option for economically viable exploration and utilization of space beyond the near-Earth environment. The bootstrapping effect of self-replication permits the utilization of vast quantities of extraterrestrial materials with only a modest initial investment of terrestrial materials.

SRS for space use must contend with an alien environment - vacuum or unusual atmospheres, zero to many gs of acceleration, radiation, temperature extremes, and so forth. Total autonomy will be more useful in space than on Earth. For symbiotic man-machine systems, man-rated life support systems are required, but because of the expense of man-rated systems it is worthwhile pursuing totally autonomous systems for early exploratory ventures. Because humans need for many reasons, to go into space in person it will ultimately be necessary to develop the required life support systems.

Possible approaches to realization of SRS. The Replicating Systems Team envisions a three-pronged approach to achieving working self-replicating systems. First, NASA should inaugurate a "top-down" program, starting with a strawman mission and defining the hierarchy of required steps for achieving that mission. Second, NASA should initiate in-house and sponsored research on enabling technologies, a "bottom-up" approach. Participation in research will keep the agency involved at the leading edge of automation technology and allow new developments to be fed into the mission design of the top-down and other NASA programs in a timely manner. The third recommended line of attack is a "middle-out" near-term hardware feasibility demonstration which will provide a focus for NASA involvement in self-replicating systems. The recommended feasibility demonstration is at the threshold of present-day technology, is extendable in a bottom-up manner to systems of greater capability and complexity, and can be decomposed in a top-down fashion to proceed from a feasibility demonstration to the fully self-replicating systems.

The top-down approach suffers from the fundamental impossibility of conceptualizing at the outset, in such an alien field of endeavor,just what the final system should be like. The bottom-up approach suffers from a lack of focus for driving it toward useful, realizable goals. Both approaches have merit and should be pursued, especially in the long run. But in the near term NASA should follow the middle-out approach and perform a feasibility demonstration which will strain the present state-of-the-art in robotics, gain NASA experience, and establish a NASA presence in state-of-the-art machine intelligence and robotics technology.

The feasibility demonstration has been conceived, however, to have three other benefits. First, when successful, it may have regenerative impact on U.S. productivity by, for example, helping to decrease the cost of robot manipulators. Second, the insights gained in performing the feasibility demonstration will be valuable in formulating a top-down mission plan for achieving extraterrestrial SRS, and in identifying valuable areas for future fundamental research and development. Third, NASA can start at the demonstration level and begin to work progressively upward toward a generalized autonomous replicating factory.

5.6.2 Top-Down Approach

The top-down approach consists first of carefully defining the overall problem, then decomposing that problem into simpler subproblems. These subproblems are, in turn, decomposed into sub-subproblems, and so on. The process continues, forming a lattice structure whose lowest tier nodes are low-level problems which are readily soluble.

Advantages and limitations. In established fields of endeavor, a top-down approach to mission and system design usually provides the most manageable solution, especially in exceedingly complex situations. Top-down structured programming in computer science is one example where this approach is beneficial. Computer software systems contain literally millions of instructions. They are, to date, mankind's most complex artifacts. Self-replicating systems will contain very complex software, in addition to being the most complex autonomous mechanical systems ever devised. For this reason, it is recommended that NASA adopt a top-down approach to the design of actual missions which employ SRS.

The top-down approach works best when there is a well established goal and a mature technology. At present it is not clear what mission employing SRS will be undertaken first. Neither is the technology mature. The mission ultimately chosen probably will depend to some extent on the outcome of basic research which has not yet been done.

Scenario for replicating systems development. To promote the achievement of self-replicating systems, NASA should identify one or more strawman missions which take advantage of self-replication. Then one of these missions should be thoroughly studied in a top-down manner.

It is recommended that the first mission to be extensively studied be a mission executed relatively close to Earth. This will minimize cost and permit human intervention if necessary. An orbiting self-replicating system or a lunar-based self-replicating system are obvious candidates. The lunar site is recommended because manufacturing engineers presently have more experience in designing industrial facilities for a planetary surface than for orbit. Traditional designs assume a surface for structural support, gravity, and maintenance of atmosphere. On the Moon only the atmosphere is absent; in orbit all three are absent.

It is recommended that the strawman mission be a Generalized Lunar Autonomous Replicating Manufacturing Facility (GLARMF). Preliminary feasibility studies of such a system have already been done (Freitas, 1980a; Freitas and Zachary, 1981; von Tiesenhausen and Darbro, 1980). The statement of work presented below is suggested for investigation of the feasibility of the strawman GLARMF mission, and is divided into five parts. All parts could be performed by one contractor; however, it would likely be beneficial to split up the work. Parts 1 and 2 probably could -best be performed by university researchers, while parts 3 through 5 might be better accomplished by one of the major aerospace companies.

Part 1: Prepare a tutorial state-of-the-art technology assessment report on autonomous manufacturing. Consider computer-aided manufacturing (CAM), computer-aided design (CAD), robotics, machine intelligence, computer vision, "telepresence" (Minsky, 1979, 1980), and other relevant fields. Separately evaluate the state-of-the-art as it exists in laboratories and in industrial practice. Determine how the state-of-the-art has progressed over time in both laboratories and in industry. Extrapolate the past and the current state-of-the-art into the future to predict when it will be feasible to construct a Generalized Lunar Autonomous Replicating Manufacturing Facility similar to that described in recent publications (Freitas, 1980a; Freitas and Zachary, 1981; von Tiesenhausen and Darbro, 1980).

Part 2: Prepare a tutorial state-of-the-art technology assessment report on nonterrestrial manufacturing. Determine how the state-of-the-art has progressed over time, both in theory and in experiment. Extrapolate the past and current state-of-the-art into the future to predict when it will be feasible to construct a Generalized Lunar Autonomous Replicating Manufacturing Facility such as that described in recent publications (Freitas, 1980a; Freitas and Zachary, 1981;von Tiesenhausen and Darbro, 1980).

Part 3: Combine the results of the technology assessment reports resulting from Part 1 on autonomous manufacturing and Part 2 on nonterrestrial manufacturing. Perform a top-down mission design for a Generalized Lunar Autonomous Replicating Manufacturing Facility. Identify those elements of the Work Breakdown Structure (WBS) which are being pursued outside NASA, but which will require additional NASA support and direction in order to achieve NASA goals. Make recommendations on how NASA should interface with the ongoing work. Identify those elements of the WBS which are unique to NASA. Make recommendations on how NASA should approach these elements.

Part 4: Perform a feasibility study for a terrestrial technology verification demonstration of a Generalized Autonomous Replicating Manufacturing Facility. Recommend one or more suitable demonstration sites. Determine what NASA in particular and the United States in general could use the facility for after the demonstration is completed. Include schedule and cost estimates (in constant dollars and real year dollars).

Part 5: Perform a feasibility study for a Generalized Lunar Automated Replicating Manufacturing Facility. Recommend one or more candidate lunar sites. Consider the construction of habitation modules and agricultural modules as output products. Compare the cost and schedule of achieving a lunar base by the use of (a) terrestrial manufacturing, (b) lunar manufacturing without replication of production facilities, and (c) lunar manufacturing with replication of production facilities. Cost estimates should be in constant dollars and real year dollars. A few suggested sources for obtaining studies of the GLARMF are listed in table 5.6.

5.6.3 Bottom-Up Approach

The bottom-up approach consists of supporting basic and applied fields related to the desired goal. Science and technology normally advance in a bottom-up fashion. Researchers build on the work of their predecessors. At any given time the problems which are soluble and present research prospects are defined by previous research which has been done and by the supporting technology which is currently available. Inventions and breakthroughs are notoriously hard to schedule in advance. It is worthwhile noting that Home sapiens, an example of an autonomous replicating manufacturing facility, was developed in a bottom-up fashion by the process of evolution.

Advantages and limitations. Occasionally, difficult goals are achieved by a concerted, directed effort. One example was sending a man to the Moon and returning him safely to Earth. Another was the Manhattan Project which produced the first atomic bomb. This approach works when the goal is clearly identified and one can determine how to achieve it. However, significant progress in science and technology is frequently made on the basis of research performed on an ad hoc speculative basis because someone is actively interested in doing that research. One of the greatest assets a nation has is the creativity and intuition of people who have devoted their lives to developing those qualities.

The top-down approach works well only when the relevant bottom-up "homework" has been done in advance. Rocketry and nuclear physics research existed long before the United States committed itself to sending a man to the Moon or developing the atomic bomb. Two good examples of how advancing technology (which was not planned to be available when the mission was designed) enhanced a mission are the high-quality TV system and the lunar rover used toward the end of the Apollo program. When people have good ideas, there should be resources available to bring those ideas to fruition.

The bottom-up approach suffers from several deficiencies. Since it is somewhat speculative in nature, some of the research will turn out to be of little use to the sponsor, though spinoffs to other fields may occur. Since bottom-up research is proposed on an ad hoc basis, careful selection is required to ensure a clear sense of direction toward the desired goal. Also, there can be some duplication of effort.

Scenario for research and development. Limitations notwithstanding, bottom-up basic and applied research is necessary to the achievement of vital and imaginative programs. Accordingly, it is recommended that NASA support moderate amounts of basic and applied research showing promise in helping to achieve NASA's goals. The mechanism that has worked fairly well (though known to have some flaws) is the publication of an Announcement of Opportunity (AO) soliciting proposals for research. These proposals are subjected to peer review, and competent ones which show some promise of payoff for NASA are funded. It is recommended that a similar mechanism be used to ensure that new ideas are factored into the mission of achieving autonomous replicative manufacturing. Otherwise, as pointed out in a recent study, unequivocal early commitment to a particular mission scenario and technology during top-down mission design will result in a mission which is using obsolete technology when it finally becomes operational.

A sample Announcement of Opportunity (AO) for SRS related basic and applied research supportive of the development of SRS technology is presented in table 5.7. It is recommended that the AO be given wide dissemination. This will allow NASA to ferret out those organizations and individuals of various persuasions, backgrounds, and in different locations who have done related research or are seriously interested in doing new research in these areas. The NASA personnel who evaluate the proposals will develop an excellent in-depth perception of the current state-of-the-art in the areas covered by the AO. This knowledge will prove invaluable when fed back to the top-down and middle-out programs.

It is recommended that the AO be distributed nationwide to the departments of industrial engineering, electrical engineering, mechanical engineering, computer science, mathematics, physics, astronomy, business, philosophy, law, and economics in colleges and universities. It is further recommended that the AO be announced in professional publications such as IEEE Spectrum; IEEE Computer; IEEE Transactions on Systems, Cybernetics, and Society; Communications of the ACM; AAAI (American Association for Artificial Intelligence) publications; SME (Society of Manufacturing Engineers) publications; Robotics Age, Industrial Robots International; Science; Science News; Byte, etc.

5.6.4 Middle-out Approach

The recommended middle-out approach consists of three stages. Briefly, in stage 1 a technology feasibility demonstration of a rudimentary self-replicating system is performed. In stage 2, stage 1 is further refined in a top-down manner to produce a less rudimentary system which operates in a less structured environment. Stage 3 consists of starting at stage 1 and doing a bottom-up synthesis of a more complex SRS.

The self-replicating system envisioned for stage 1 is a computer connected to one or more manipulators. Under control of the computer, the manipulator(s) will assemble another computer and another set of manipulator(s) from well-defined subassemblies. Examples of these subassemblies are printed circuit cards for the computer and individual joints or limb sections for the manipulator(s). This approach to self-replication is inspired by the von Neumann "kinematic model" as described in section 5.2.

In stage 2, the subassemblies would begin to be assembled from still smaller sub-subassemblies such as integrated circuits, resistors, motors, bearings, shafts, and gears. This stage can proceed for quite some time as the techniques for assembling each subassembly from sub-subassemblies are developed and implemented one by one. By the time stage 2 is complete, there will be extensive crossfertilization taking place between industry and the feasibility demonstration Indeed, accomplishment of stage 2 will mean that robots can be assembled from parts by other robots. As discussed in sections 5.4 and 5.5,this will have a profound impact on U.S. industry.

Stage 3 is the final link in achieving an autonomous self-replicating manufacturing facility. In stage 3 the manipulators, which have, in stages 1 and 2, been assembling more robots, are used to build the machines which make the parts. For example, the manipulators could assemble a printed circuit board manufacturing machine or a gear manufacturing machine. The problem of closure, discussed at length in section 5.3.6, becomes a major practical issue at this point. One must be careful that as one adds more and more machines the total number of different parts required is eventually produced by the total population of machines.

Advantages. The middle-out approach has a number of important advantages. In the long run it will replace neither the top-down nor the bottom-up methodologies. It does, however, provide a place to start on the practical realization of SRS.

The middle-out approach begins with the feasibility demonstration and then proceeds in a top-down and a bottom-up fashion. The feasibility demonstration alone will produce useful output - the automated production of robots. The expenditure required for the feasibility demonstration is tiny compared to the expenditure required before either the top-down or the bottom-up approach begins to show useful output. The middle-out approach can then be continued at whatever level of support seems appropriate and will produce useful spinoffs for industry as it progresses.

One of the chief advantages to NASA of the feasibility demonstration is that it can begin immediately. Working on the feasibility demonstration will provide NASA with valuable insights into practical problems associated with self-replicating systems. These insights will greatly increase the efficiency with which NASA can pursue both the top-down and the bottom-up approaches. The feasibility demonstration will be a valuable learning tool for both NASA and the industrial community.

As has been previously stated in this report, achievement of robot production of robots will decrease the cost of robots. This will directly benefit U.S. productivity and indirectly benefit NASA by lowering the cost of manufactured goods. Another valuable characteristic of the feasibility demonstration is that it will produce a visible output a functioning autonomous self-replicating system (albeit a rudimentary one). In a field which is as foreign to most people as autonomous SRS, this will lend valuable credibility to the plans to produce more complex autonomous systems in space.

Limitations. The chief limitation of the middle-out approach is that it will not, of itself, produce an autonomous self-replicating system suitable for NASA's needs in space. The direction provided by the top-down approach is also needed. Also the creativity of the bottom-up approach is necessary to provide the needed adaptations to the space environment, such as designs and processes optimized for the use of extraterrestrial materials. Another disadvantage of the middle-out approach is that it will consume resources which could otherwise be devoted to the top-down and bottom-up methodologies. However, the overall efficiency should be greatest if a balance is maintained among all three approaches.

As simple as it sounds, the team estimates, on the basis of its discussions with industry and research community representatives, that it would require about 5 years and $5-50 million (1980 dollars) to accomplish the feasibility demonstration proposed below. The major difficulties include the following:


 * Assembly by robot is a difficult task at present, and final assembly is one of the more difficult forms of assembly.
 * Present-day robot manipulators are built using hand labor. They are not designed for easy automated assembly. American Robot Corporation is reported to be planning on the automated assembly of robots beginning in 1981 (Industrial Robots International, 1980). However, these robot manipulators are quite small (5 lb load capacity), and "Gallaher's forecasts of small robot acceptance seem highly optimistic as do his own production plans and pricing." The Japanese have been far more aggressive in this area (IAF Conference, 1980).
 * Present-day robot manipulators are rather weak for their weight. Care must be exercised to ensure that the subassemblies are light enough for the robot manipulators to be able to manipulate them - or, alternatively, to ensure that the robot manipulator is strong enough to be able to manipulate the subassemblies.

These problems are by no means insurmountable. However, considerable reengineering of robot manipulators will be required to facilitate their assembly by similar robot manipulators. Likewise, the packaging of the computer will require some re-engineering for easy assembly by a robot manipulator.

Scenario for replicating systems demonstration. We now present a more detailed description of the proposed demonstration scenario for SRS. The demonstration begins with a parts depot stocked with enough subassemblies for the production of two robot manipulators and their associated computer systems. One complete, operating robot, Robot i, is also present. It will construct Robot 2 which will, in turn, construct Robot 3, thus passing the "Fertility Test" (sec. 5.3.3). This arrangement is shown schematically in figure 5.28.



Robot 1 begins its labors by obtaining, one at a time, the subassemblies for the base (which doubles as the electronics card cage assembly) of Robot 2 from the parts depot. Robot 1 assembles the base, computer, and servo controls for Robot 2. Then, one at a time, Robot 1 obtains the subassemblies for the manipulator arms of Robot 2 and constructs the arms of Robot 2 from them.

When Robot 2 has been completely assembled, Robot 1 plugs in the power cord of Robot 2. Robot 1 then obtains a blank diskette (a removable mass memory device for computers) from the parts depot, inserts the diskette into its own computer, copies its software onto the diskette, and then removes the diskette from its own computer. Reproduction is complete when Robot 1 turns on the power to Robot 2, inserts the diskette (which now has a copy of the operating software on it) into Robot 2's computer, and then pushes the start button on the computer. From then on, Robot 2 is autonomous.

It should be noted that some additional complexity has been introduced into the demonstration by explicitly transferring the instructions from one generation of robot to the next by physical movement of a recording medium. This strategy was employed to make it clear that the generations are truly autonomous.

One of the ground rules of a demonstration such as this should be that all interaction between the robots be explicit and visible to a human observer. If the computers of the various robots were electrically interconnected the psychological impact on the observer would be more along the lines of a single system which was expanding itself, rather than producing distinct offspring. In addition, the demonstration as described should have an especially significant impact on anyone who has ever inserted a diskette into an inert computer and activated it by "booting it up."

The demonstration then proceeds by having Robot 2 construct and activate Robot 3. Robot 2 obtains the parts from Robot I, who obtains them in turn from the parts depot and passes them along bucket-brigade style, according to its stored post-replication instructions. Af