« PreviousContinue »
NASA End-to-End Data Management System. The NASA end-to-end management system outlined in this report must undergo considerable planning and detail before it can become a reality. A systems engineering group consisting of computer scientists and hardware experts is needed now to achieve an effective system concept and implementation.
Knowledge Centers. The concept of knowledge centers must be explored carefully from a technical level to determine how they are to be achieved. Careful consideration will be required to develop or obtain appropriate archival and data base management systems. Insufficient attention is being placed on this aspect of NEEDS.
Knowledge Base Systems. Research support is required for work in this area. Emphasis should be placed on enhancing relational data base systems so that they can be used in conjunction with problem solving systems needed to achieve knowledge base system capabilities. A knowledge base system can be achieved, and is necessary for NEEDS. Again, insufficient attention is being placed on such systems.
8. Man-Machine Systems Technology
This section deals with the three major components of any advanced man-machine system: modeling human control processes, the design of interfaces between human and intelligent computer systems, and the design of the manipulators themselves.
The major deficiency in the application of machine intelligence and robotics to the special problems of NASA is a lack of knowledge about how to design effective man-machine systems. This deficiency is fundamental and is based on a lack of knowledge of human processes of machine control and of the interface. For example, teleoperators are understood neither at the level of the human control processes nor at the level of determining an appropriate design of manipulation and the proper design of the information interface between human and teleoperators.
There does exist considerable knowledge about each of the fields that contribute to their problems. Cognitive psychology has developed considerable expertise and knowledge about the structures of human information processing, most especially those of perception, language, and memory. Workers in control theory have developed sophisticated procedures.
Human-computer interaction has been widely studied. The weaknesses lie in the lack of useful predictive models of the interaction of these components.
Despite the fact that human-machine interaction is critical to the success of almost all of NASA's missions, NASA's present organizational structure does not appear to accommodate research on man-machine systems or man-computer interaction required. NASA Life Science programs (under Office of Space Science) have concentrated on medical, physiological, botanical, bacteriological, and biochemical disciplines. OAST has sponsored some man-machine research, but primarily as related to aeronautics. The NASA centers have done more man-machine research on an ad hoc basis, as required by the project offices. Thus, human information processing, man-computer cooperation, and man-machine control basic research has tended to "fall between the cracks."
The result is that in current and planned missions there is confusion about what mechanisms are most appropriate for communication in control and feedback of information to human operators, and what constitute appropriate tasks for humans and for machines. So far, the ambiguous status of the human-machine systems research has not led to any grave difficulties, primarily because the conservative engineering philosophy of NASA helps avoid major difficulties. The lack does severely limit applications, however.
Fundamental improvements in human-machine interaction will come about only if NASA leads the way, supporting basic research directed at the fundamental problems, developing applied laboratories, developing new conceptualizations and new techniques. This work must be mission independent, developed from broadly based fundamental research and development programs that are not subject to the complex limitations posed by mission-oriented studies. There must be better means for life science and technology organizational components to interact in bringing a more rigorous focus on these crucial long-range research problems.
8.2 Human Information Processing
Human information processing is the study of the psychological mechanisms underlying mental functioning. Memory, problem solving, language, perception, thinking - these are some of the major areas studied. In the past decade there have been sufficient systematic advances in our knowledge that these areas now constitute perhaps the best understood problems in contemporary psychology. Studies of attention are of special importance to problems faced by NASA. Humans have limited mental resources, and the deployment of these resources constitutes an important part of behavior. The limitation appears to apply primarily to conscious control.
Tasks that require conscious decision-making or control can suffer in times of stress or when other tasks must be performed or thought about simultaneously. When several tasks simultaneously demand a share of conscious resources, deterioration of performance results. Tasks that are learned well enough that they appear "automated" seem to suffer little as a result of other activity.
Despite the relative amount of knowledge about human processing mechanisms and control structures, we know surprisingly little about aspects that are relevant to the problems faced by NASA. We do not know enough about the nature of conscious and subconscious-control mechanisms. We do not know enough about the various modes of operation of the human. We know very little about the human's ability to interact with and control the environment. Almost all our knowledge deals with the processing of arriving information, or the operation of the human as an element of a control structure. This leaves unanswered much of importance. The human has two cortical hemispheres, each one appearing to be specialized for different types of processing. One hemisphere appears to be serial, the other more parallel or distributed. We have just begun to explore the implications of these differences; exactly how they apply to control issues is not understood, although there are obvious implications.
This area of knowledge about the human has many potential applications for NASA. On the spacecraft, at mission control, onsite during a mission, all these situations require different aspects of human capability. We find no evidence that NASA is engaged in systematic study of these issues. Yet they are critical to the success of NASA's missions, the more so as missions become longer, more complex, with space repair, manufacture, and mining as possible tasks.
8.3 Human-Machine Control Processes
Whether humans are physically present in space or not, their intelligent control and supervision are fundamental requirements for any space mission within this century. Their sensory, cognitive, and motor-control systems can play useful roles in combination with machine intelligence. The combination of human and computer has potential for more capability and reliability than either by itself. But realization of this synthesis requires much progress at the level of basic research and application.
A primitive discipline and art of human-machine systems does now exist, although it is unevenly developed. There are a small number of texts and several scientific journals. There are regular workshops and annual meetings. In such areas as the
performance of the pilot and air traffic controller in aviation, the work has been sufficiently successful that aircraft manufacturers and government regulators determine their hardware and procedures to a significant degree from research findings. But the most sophisticated of this research and the most successful empirical applications have been devoted to situation, where the human is in continuous control of the system. As the computer becomes more capable of intelligent operation, the human operator becomes more like a "manager" or a "supervisor" than an active in-the-loop controller. This new "supervisory control" mode of operation is not well understood. Ames and Langley both have in-house and university research programs to study these new roles in the context of aviation. This is a start, but more needs to be done, especially directed towards the particular problems faced by space programs.
8.4 Human Interactions For Ground-Based Missions
A large fraction of NASA's budget is spent on an extensive system for monitoring, controlling, and processing data from a large number of Earth-orbital and deep-space vehicles. The Study Group severely questions much of this work. There has not been sufficient research done on the proper balance between human and machine judgment and analysis, not proper studies of the appropriate balance between decisions made in space and on the ground at mission control. While the new color-graphic computer based displays offer much possibility, their appropriate display formats are not well understood. The role of intelligent programs is significantly underestimated.
Ever since the development of remote manipulators for nuclear applications in the early 1950s, it has been clear that teleoperators can be used to extend the human's vision, mobility, and manipulation capability into space, undersea, and difficult environments. Nonetheless, compared to the magnitude of the potential saving over sending man into space to sense and manipulate, there has been little developmental work on the control factors of teleoperators. (In similar fashion, there has been surprisingly little development of the manipulators themselves, but this is covered in a different part of this report.) When continuous remote control is not possible because of signal transmission time delays, restrictive "move and wait" strategies are required for remote operation. This can lend to awkwardness and instability, especially when force or touch feedback is used. It is clear that to perform large space construction tasks or planetary mining, etc., in other than near-Earth orbit, this type of control is intolerable.
The Study Group finds that there has been surprisingly little advancement in manipulator development since the 1960s, though recently some significant theoretical contributions to kinematics and control of manipulators are evident in both U.S. and Soviet literature. Current manipulators do not have the reach, precision, or sense ability required for space assembly. End effectors are awkward and tool changing is slow. Even use of the human operator in a supervisory mode (with a computer doing much of the control), requires close contact with the task via television and force-reflecting sensing mechanisms. Manipulation dexterity must be understood better at a fundamental level.
Thus, assuming semiautomatic assembly, where the human plays a supervisory or decision-making role, two things are needed: development of intelligent computer control systems, and the understanding of the role of the human in this mode of operation. There has been insufficient research to understand the proper interface that should exist when the human plays this higher-order role in the feedback system.
NASA sponsored research dealing with capabilities of the human-machine interaction are bound to lead to important spin-offs. Increased understanding of sensory and control mechanisms will be important for the development of sensory prostheses, for development of new systems to aid in the control and management of complex tasks, and perhaps systems capable of expanding our cognitive abilities by cooperative interaction with machines.
Research on man-machine questions should have direct application to various non-NASA needs such as unmanned mining, deep ocean exploration and construction, disposal of nuclear waste, and intracorporeal surgery performed remotely using optical fiber bundle devices. New techniques of industrial automation are a possible outgrowth of such research.
NASA must reassess the role of the human in the control of sophisticated systems. Low level, detailed control will probably best be done by intelligent computers, giving the humans higher level, more complex decision making and administrative responsibilities.
NASA should develop a strong research program on human information processing, on man-machine control processes,
and on the interface issue. There is special need for study of situations with high data rates, with a need for rapid decisions, and with stress. There should be direct interrelationships between NASA's mission needs and the research programs. Formal ties with the university and industrial research communities would be useful, with collaborative research being potentially of great value. A scientific visitors' program could educate NASA scientists to existing research; young university personnel would become educated about NASA's particular problems.
Note that many of the problems faced by NASA occur in rich, complex environments. University research laboratories are unlikely to have the facilities to simulate these conditions. Accordingly, if NASA laboratories were made available to researchers from university settings, with sufficient time and resources, it might be possible simultaneously to increase the level of basic understanding of problems dealing with manmachine interface, and also to get direct results relevant to NASA. Thus, experiments on simultaneous attention, or on performance under stress, or on the control of teleoperators, using NASA simulators can be expected to make it possible for new kinds of phenomena to be studied. NASA has the facilities, but has not used them for general development. Universities have the technical expertise, but lack the facilities to do research relevant to the real needs of NASA.
In addition, the Study Group recommends:
1. That research in the areas of man-computer cooperation and man-machine communication and control be accelerated, in view of the long-range critical need for basic understanding of these problems. This is in lieu of supporting such research primarily on an ad hoc and mission-oriented basis.
2. That NASA organizational entities representing life sciences and the technological disciplines of computers and control develop better cooperative mechanisms and more coherent research programs in man-machine interaction to avoid the "falling between the cracks" problem.
3. That future NASA missions realize the full potential of teleoperators by developing improved means for "supervisory control" of robotic teleoperators. Thus the human operator on Earth can benefit from what the teleoperator senses, and can intermittently reprogram its computer as necessary. In this way the advantages of human intelligence in space may be had without the costs and risks of bodily presence there.
9. Digital Communication Technology
There is an aspect of NASA activity that did not receive much attention at any of the workshop meetings. This involves the transfer of information among a complex, geographically and institutionally disparate set of groups that need to exchange messages, ideas, requirements, documents, to keep informed, plan activities, and arrive at decisions quickly. The clerical infrastructure to support this network of activity, not counting the information users and generators, managers, engineers, and scientists at the nodes, must account for approximately 15% of NASA's budget for both inhouse and contractor personnel. If total personnel costs are 2/3 of the total budget, then the total costs for the mechanics of this information exchange is several hundred million dollars per year. A computer-based communication system can make significant improvements in the quality of information transfer, and probably increase the productivity of the information exchange infrastructure.
The implementation of such a system would not be predicated on new developments in artificial intelligence. It would use tools that are common practice at AI nodes of the ARPA network and are part of the developing technology of digital information and word processing. Once such a development were carried out, it would provide the data base that could take advantage of sophisticated techniques of information retrieval, semantic search, and decision making as they became available. Costs can be estimated accurately from systems at those AI sites on the ARPA network.
The functions to be carried out are the Directory and File management facilities described in the TENEX Executive manual and programs like SENDMESSAGE, MESSAGE, TV EDIT, and BULLETIN BOARD. These would operate interactively. Programs like SPELL and PUB would be offered. Teleconferencing facilities and input of documents with OCR devices could be implemented. These services offer mail to distribution lists, creation and editing of documents, spelling checking, and publication formatting, with book quality print, arbitrary fonts and graphics, with quick turnaround. Documents would be instantly available for online access and continuous updating.
In addition to the cost savings, which would probably be large, there would be the following:
Conclusions and Recommendations
We believe that NASA should institute a vigorous and longrange program to incorporate and keep pace with state-of-theart developments in computer technology, both in its spaceborne and its ground-based computer systems; and to ensure that advances, tailored to NASA's mission, continue to be made in machine intelligence and robotics. Such advances will not occur of their own accord. Many NASA requirements in computer architecture and subsystem design will in turn have a stimulating effect on the American computer and microprocessor industry, which now faces an extremely strong challenge by foreign competition. We believe that an agency such as NASA, which is devoted to the sophisticated acquisition and analysis of data, must play a much more vigorous role in the design and acquisition of data processing systems than has been its practice in the past.
These findings are supported by the recommendations independently arrived at by the Space Science Board of the National Academy of Sciences":
From experience with mission operations on previous space missions, we anticipate that there will be even greater demands on data acquisition, processing, and storage; on mission coordination; and on interaction with the spacecraft and scientific experiments. The complex nature of mission operations and the long time scale required to prepare, certify, and transmit routine commands in previous missions indicates that substantial changes will be necessary. We believe that significant technical and managerial advances must be made in anticipation of future planetary missions, in order to provide reliable, more efficient, and lower cost systems for operation of the spacecraft and scientific instruments.
The testing of these systems on the ground as operational units including the participation of science teams should be carried out well before the mission. These tests should include the operation with possible failure modes. These approaches will be more important in the future when extensive coordination must be obtained by use of more intelligent or autonomous control systems. The
Strategy for Exploration of the Inner Planets: 1977-1987, Committee on Planetary and Lunar Exploration, Space Science Board, Assembly of Mathematical and Physical Sciences, National Research Council, National Academy of Sciences, Washington, D.C., 1978.
choice of onboard preprocessing versus earth-based processing and the utility of block telemetry formatting and distributive data handling and control subsystems will require assessment. In the past, computing facilities and command and dataprocessing software were not always efficient, and early attention was not given to overall system design in laying out missions. Further, experience with past and current spaceflight missions has shown that complicated systems with higher levels of intelligence are difficult to handle without substantial experience.
We are apprehensive about recommending that radical new approaches be utilized without further study; nonetheless, it appears that some significant changes must be considered. Recognizing that mission operations is the key to the success of any complicated undertaking, we therefore recommend that an assessment of mission operations, including spacecraft control and scientific instrument and data management and the design and management of software control systems, be studied by the Agency at the earliest possible time and the evaluation be presented to the Committee.
The Federal Data Processing Reorganization Project has indicated serious failings in virtually all government agencies in the utilization of modern computer technology. While the National Science Foundation and the Advanced Research Project Agency (ARPA) of the Department of Defense continue to support some work in machine intelligence and robotics, this work, especially that supported by ARPA, is becoming more and more mission-oriented. The amount of fundamental research supported by these agencies in machine intelligence and robotics is quite small. Because of its mission, NASA is uniquely suitable as the lead civilian agency in the federal government for the development of frontier technology in computer science, machine intelligence, and robotics. NASA's general engineering competence and ability to carry out complex missions is widely noted and admired. These are just the capabilities needed by any federal agency designated to develop these fields. Although we are hardly experts on federal budgetary deliberations, it seems to us possible that incremental funds might be made available to NASA, over and above the usual NASA budget, if NASA were to make a compelling case for becoming the lead agency in the development of frontier technology in computer science and applications.