Large-area systems such as large space antennas, satellite power systems, and space stations require large-scale and complex construction facilities in space (Figures 4-5 and 4-6). Relatively small systems, up to 100 m in extent, may be deployable and can be transported into orbit with one Shuttle load. For intermediate systems of several hundred meters in extent, it becomes practical to shuttle the structural elements into space and assemble them on site (Figure 4-7). Very large systems require heavy-lift launch vehicles which will bring bulk material to a construction platform (Figure 4-8), where the structural components are manufactured using specialized automated machines. The structural elements can be handled by teleoperated or self-actuating cranes and manipulators which bring the components into place and join them (Figure 4-9). Free-flying robots will transport the structural entities between the Shuttle or the fabrication site and their final destination and connect them. These operations require a sophisticated general-purpose Figure 4-5. Large space systems require robot and automation technology for fabrication, assembly, and construction in space. Figure 4-7. Construction of a space station. Bulk material is brought by the Shuttle. Structural elements are fabricated at the construction facility and then assembled by remotely controlled manipulators. Figure 4-6. Large space antennas are erected with the help of a space-based construction platform. The Shuttle brings the structural elements to the platform, where automatic manipulator modules under remote control perform the assembly. Figure 4-8. Complex construction facility in space with automatic beam builders, cranes, manipulators, etc., is served by the Shuttle. handling capability. In addition to transporting structural elements, the robot must have manipulators to handle them, and work with them and on them. Large structural subsystems must be moved from place to place and attached to each other. This usually requires rendezvous, stationkeeping, and docking operations at several points simultaneously and with high precision - a problem area still not investigated for zero gravity. Automated "smart" tools would also be required by astronauts to perform specialized local tasks. These robot systems could be controlled remotely as teleoperator devices, or they could be under supervisory control with intermittent human operator involvement. Astronauts in space or human operators on Earth will need the tools to accomplish the envisioned programs. The technology for in-space assembly and construction will provide the foundation for the development of these space-age tools. After the system has been constructed, its subsequent operation will require service functions that should be performed by free-flying robots or by robots attached to the structure. The functions which such a robot should be able to perform include calibration, checkout, data retrieval, resupply, maintenance, repair, replacement of parts, cargo and crew transfer, and recovery of spacecraft. During and after construction, there should be a robot on standby for rescue operations. An astronaut drifting into space could be brought back by a free-flying robot. Such devices could also be on stand-by alert on the ground. The delivery systems for these rescue robots need not be man-rated. They can deliver expendable life support systems or encapsulate the astronaut in a life support environment for return to a shuttle, space station, or Earth. They could also perform first-aid functions. Another phase of space industrialization calls for a lunar or asteroidal base. After a surface site survey with robot (rover) vehicles, an automated precursor processor system could be placed on the Moon or the asteroid. This system would collect solar energy and use it in experimental, automated physical/ chemical processes for extracting volatiles, oxygen, metals, and glass from lunar soil delivered by automated rovers (Figure 4-10). The products would be stored, slowly building up stockpiles in preparation for construction. The lunar or asteroidal base would be built using automated equipment and robots as in Earth orbit. After construction, general-purpose robot devices would be necessary for maintenance and repair operations. In addition, the base would use industrial automation (qualified for operation in space) or a sort generally similar to those employed on Earth for similar tasks. Figure 4-9. Space construction of large antenna systems with automated tools, teleoperated manipulators, and free-flying robots. Figure 4-10. Automated material processors on the lunar surface are serviced by robot vehicles with raw lunar soil. Section V Technological Opportunities A. Trends in Technology Machine intelligence and robotics are not only relevant but essential to the entire range of future NASA activities. Content analysis of Earth orbital and planetary spacecraft results is merely one application. Other applications exist: in mission operations, in spacecraft crisis management, in large constructions in Earth orbit or on the Moon, and in mining in the lunar or asteroidal environments. These last applications are probably at least a decade into the future, but some essential preparations for them would seem prudent. These preparations might include the development of teleoperators, manipulative devices which are connected via a radio feedback loop with a human being, so that, for example, when the human on the Earth stretches out his hand, the mechanical hand of the teleoperator in Earth orbit extends likewise; or when the human turns his head to the left, the teleoperator's cameras turn to the left so that the human controller can see the corresponding field of view. Where the light travel times are on the order of a tenth of a second or less, the teleoperator mode can work readily. For repetitive operations, such as girder construction and quality control in large space structures, automation and machine intelligence will play a major role in any efficient and cost-effective design. In planetary exploration in the outer solar system, the light-travel times range from tens of minutes to many hours. As a result, it is often useless for a spacecraft in trouble to radio the Earth for instructions. In many cases, the instructions will have arrived too late to avoid catastrophe. Thus, the Viking spacecraft during entry had to be able to monitor and adjust angle of attack, atmospheric drag, parachute deployment, and retro-rocket firing. Roving vehicles on Mars, Titan, and the Galilean satellites of Jupiter will have to know how to avoid obstacles during terrain traverses and how not to fall down crevasses. The development of modern scientific spacecraft necessarily involves pushing back the frontiers of machine intelligence. In our opinion, machine intelligence and robotics is one of the few areas where spinoff justifications for NASA activities are valid. In most such arguments, socially useful applications, such as cardiac pacemakers, are used to justify very large NASA expenditures directed toward quite different objectives. But it is easy to see that the direct development of the application, in this case the pacemaker, could have been accomplished at a tiny fraction of the cost of the activity which it is used to justify the Apollo program, say. However, because there is so little development in machine intelligence and robotics elsewhere in the government (or in the private sector), spinoff arguments for NASA involvement in such activities seem to have some substantial validity. In the long term, practical terrestrial applications might include undersea mineral prospecting and mining, conventional mining (of coal, for example), automated assembly of devices, microsurgery and robotics prosthetic devices, the safe operation of nuclear power plants or other industries which have side effects potentially dangerous for human health, and household robots. A further discussion of future NASA applications of machine intelligence and robotics, and possible spinoff of these activities, is given in the supporting documentation. With the development of integrated circuits, microprocessors, and silicon chip technology, the capabilities of computers have been growing at an astonishing rate. Figures 5-1 through 5-4 provide estimates of recent past and projected future developments. By such criteria as memory storage, power efficiency, size and cost, the figures of merit of computer systems have been doubling approximately every year. This implies a thousand-fold improvement in a decade. In another decade the processor and memory (four million words) of the IBM An interesting possible application of general purpose robotics technology is provided by the nuclear accident at the Three Mile Island reactor facility near Harrisburg, Pennsylvania in March/April 1979. The buildup of a high pressure tritium bubble had as one possible solution the turning of a valve in a chamber under two meters of water impregnated with very high radiation fluxes. This is an extremely difficult environment for humans, but a plausible one for advanced multipurpose robots. The stationing of such robots as safety devices in nuclear power plants is one conceivable objective of the development of robotics technology. Generally, such multipurpose robots might be stationed in all appropriate industrial facilities where significant hazards to employee or public health or to the facility itself exists. Shortly after the Three Mile Island reactor accident the operating company began recruiting "jumpers," individuals of short stature willing, for comparatively high wages, to subject themselves to high radiation doses thought inappropriate for permanent reactor technicians (New York Times, July 16, 1979, page 1). The functions are often no more difficult than turning a bolt, but in a radiation environment of tens of rems per hour. There would appear to be strong humanitarian reasons for employing small multipurpose self-propelled robots for this function, as well as to redesign nuclear power plants to make much fuller use of the capabilities of machine intelligence. The competent use of machine intelligence and robotics is an important component of all recently proposed additional energy sources for example, mining and processing shale and coal. Figure 5-2. Active devices technology. The number of active components per cubic centimeter is doubling every 1-1/8 years, whereas the average cost per logic gate is halving every 2-1/2 years. Figure 5-4. Computer systems technology. The average increase of computer speed is doubling every 1-1/2 years, whereas the failure rate is halving every 2-3/4 years. 10-15 10-16 1990 10-13 FAILURES/bit X sec 100 λ(μm), d(μm), Mbits/sec 370/168 will probably be houseable in a cube about five centimeters on a side (although computer architecture different from that of the IBM 370/168 will probably be considered desirable). It is difficult to think of another area of recent technology which has undergone so many spectacular improvements in so short a period of time. This steep rate of change in computer technology is one major factor in the obsolescence of NASA computer systems. New systems are being developed so fast that project scientists and engineers, mission directors, and other NASA officials have difficulty discovering what the latest advances are, much less incorporating them into spacecraft-mission or groundoperations design. Another problem is the competition between short-term and long-term objectives in the light of the NASA budget cycle. Major funding is given for specific missions. There is a high premium on the success of individual missions. The safest course always seems to be to use a computer system which has already been tested successfully in some previous mission. But most missions have five- to ten-year lead times. The net result is that the same obsolete systems may be flown for a decade or more. This trend can be seen in areas other than computer technology, as, for example, in the NASA reliance in lunar and planetary exploration for 15 years on vidicon technology, well into a period when commercial manufacturers were no longer producing the vidicon systems and NASA was relying on previously stockpiled devices. This has been the case since 1962. Only with the Galileo mission, in 1984, will more advanced and photometrically accurate charged-coupled device systems be employed. The problem is much more severe when it applies to a field undergoing such dramatic advances as computer technology. The management dynamics can be understood, but it is nevertheless distressing to discover that an agency as dependent on high technology as NASA, an organization identified in the public eye with effective use of computer technology, has been so sluggish in adopting advances made more than a decade earlier, and even slower in promoting or encouraging new advances in robotics and machine intelligence. The general technological practice of adopting for long periods of time the first system which works at all rather than developing the optimal, most cost-effective system has been amply documented. This phenomenon is by no means restricted to NASA. The need to handle radioactive substances led many years ago to the development of rudimentary teleoperators. At first progress was rapid, with force reflecting, two-fingered models appearing in the early 1950s. But this 7Simon, Herbert A., The New Science of Management Decision, revised edition, Prentice-Hall, Inc., Englewood Cliffs, NJ, 1977. development all but stopped when progress was sufficient to make the handling of radioactive materials possible - rather than easy, or economical, or completely safe. This occurred in part because the nuclear industry, like NASA, became mission-oriented at this time. Since then, the development of computer-controlled manipulators has proceeded slowly on relatively sparse funding, and there has been little drive to understand in a general and scientific way the nature of manipulation. Major advances seem similarly stalled and likewise entirely feasible in such areas as locomotion research, automated assembly, self-programming, obstacle avoidance during planetary landfall, and the development of spacecraft crisis analysis systems. B. Relevant Technologies The principal activity of the Study Group during its existence was to identify machine intelligence and robotics technologies that are highly relevant to NASA and the success of its future programs. Each Study Group workshop had one or more of these topics as the foci of interest. Appendix A gives a complete list of topics covered at each of the workshops. In this section we provide a summary of the discussions of the topics considered by the Study Group. 1. Robotics Technology Robotics and machine intelligence have in the past played surprisingly small roles in NASA space programs and research and development. Yet these areas will become increasingly more important as the emphasis shifts from exploration missions to missions involving space utilization and industrialization and the fabrication and assembly of space structures. The high cost of placing people in space suggests that the use of robots might be the method of choice long before robotics become practical on Earth. The uses of robotics can be broadly grouped into manipulators and intelligent planetary explorers. There already exist automatic vision and manipulation techniques that could be developed into practical systems for automatic inspection and assembly of components. Parts could be marked to allow simple visual tracking programs to roughly position them, while force-sensing manipulators could mate the components. Where large structures, designed from standard sets of component parts and assembled in regular patterns are concerned, manipulators could perform reliable, accurate, repetitive operations which would be difficult for a human, in space, to do. Intelligent robot explorers will become imperative, if sophisticated large-scale interplanetary exploration is to become a reality. The round-trip communication delay time |