« PreviousContinue »
Section IV Applications of Machine Intelligence and Robotics in the Space Program
The space program is at the threshold of a new era that may be distinguished by a highly capable space transportation system. In the 1980s, the Space Shuttle and its adjuncts will enable increased activities in the scientific exploration of the universe and a broadened approach to global service undertakings in space. The first steps toward utilizing the space environment for industrial and commercial ventures will become possible and can trigger requirements for more advanced space transportation systems in the 1990s. This will enable expanded space industrial activities and, by the end of this century, could lead to Satellite Power Systems for solar energy production, to lunar or asteroidal bases for extracting and processing material resources, and to manned space stations for commercial processing and manufacturing in space. A major objective for NASA is to develop the enabling technology and to reduce the costs for operating such large-scale systems during the next two decades. On examining potential NASA missions in this time frame we expect that machine intelligence and robotics technology will be a vital contributor to the costeffective implementation and operation of the required systems. In some areas, it will make the system feasible, not only for technological reasons, but also in terms of commercial acceptability and affordability.
humans. By means of telecommunication, humans can activate and control systems at remote places. They can perform tasks even as far away as the planets. During the 1960s, this became known as teleoperation. Teleoperators are man-machine systems that augment and extend human sensory, manipulative, and cognitive abilities to remote places. In this context, the term robot can then be applied to the remote system of a teleoperator, if it has at least some degree of autonomous sensing, decision-making, and/or action capability. The concept of teleoperation has profound significance in the space program. Because of the large distances involved, almost all space missions fall within the teleoperator definition; and, because of the resultant communication delay for many missions, the remote system requires autonomous capabilities for effective operation. The savings of operations time for deep space missions can become tremendous, if the remote system is able to accomplish its tasks with minimum ground support. For example, it has been estimated that a Mars roving vehicle would be operative only 4 percent of the time in a so-called move-and-wait mode of operation. With adequate robot technology, it should be operative at least 80 percent of the time.
During the next two decades, the space program will shift at least some emphasis from exploration to utilization of the space environment. It is expected that this shift will be accompanied by a large increase in requirements for system operations in space and on the ground, calling for general-purpose automation (robotics) and specialized automation. What operations, tasks, and functions must be automated, and to what degree, to accomplish the NASA objectives with the most cost-effective systems?
NASA saw the need to examine the civilian role of the U.S. space program during the last quarter of this century. A series of planning studies and workshops was initiated with the Outlook for Space Study in 1974, which included a comprehensive forecast of space technology for 1980-2000. In a subsequent NASA/OAST Space Theme Workshop, the technology forecasts were applied to three broad mission themes: space exploration, global services, and space industrialization. Based on the derived requirements for cost-effective space mission operations, five new directions were identified for developments in computer systems, machine intelligence and robotics: (1) automated operations aimed at a tenfold reduction in mission support costs; (2) precision pointing and control; (3) efficient data acquisition to permit a tenfold increase in information collection needed for global coverage; (4) real-time data management; and (5) low-cost data distribution to allow a thousand-fold increase in information availability and space-systems effectiveness. The machine intelligence and automation technologies for data acquisition, data processing, information extraction, and decision making emerge here as the major drivers in each area and call for their systematic development. In addition, for certain areas such as automated operations in space, the mechanical technologies directed at
B. Robots and Automation in
Whereas mechanical power provides physical amplification and computers provide intellectual amplification, telecommunication provides amplification of the space accessible to
SExcerpted from New Luster for Space Robots and Automation by
Ewald Heer, Astronautics & Aeronautics, Volume 16, No 9, pp 48-60, September 1978.
materials and objects acquisition, handling, and assembly must also be further developed; robots doing construction work in Earth orbit or on the lunar surface will need manipulative and locomotion devices to perform the necessary transport and handling operations.
C. Future Applications
In space applications, robots may take on many forms. None looks like the popular science fiction conception of a mechanical man. Their appearance follows strictly functional lines, satisfying the requirements of the mission objectives to be accomplished. The discussion which follows briefly presents mission categories, mission objectives, and system characteristics pertinent to space robot and automation technology. Estimates of technology development efforts to automate system functions are given in Table 4-1.
1. Space Exploration
Space exploration robots may be exploring space from Earth orbit as orbiting telescopes, or they may be planetary flyby and/or orbiting spacecraft like the Mariner and Pioneer families. They may be stationary landers with or without manipulators like the Surveyor and the Viking spacecraft, or they may be wheeled like the Lunakhod and the proposed Mars rovers. Others may be penetrators, flyers, or balloons, and some may bring science samples back to Earth (Figures 4-1 4-3). All can acquire scientific and engineering data
Figure 4-2. Mars surface robot will operate for 2 years and travel about
1000 km performing experiments automatically and sending the scientific information back to Earth.
Figure 4-1. Galileo spacecraft navigates between Jupiter and Galilean
satellites in rendering. After sending a probe into the jovian
Figure 4-3. Artist's concept of a Mars surface scientific processing and
sample return facility. Airplanes transport samples into the vicinity of the processing station. Tethered small rovers then bring the samples to the station for appropriate analysis and return to Earth.
using their sensors, process the data with their computers, plan and make decisions, and send some of the data back to Earth. Some robots are, in addition, able to propel themselves safely to different places and to use actuators, manipulators, and tools to acquire samples, prepare them, experiment in situ with them, or bring them back to Earth.
information is short-lived except for possible historical reasons. The value of information of disasters such as forest fires is of comparably short duration. The demand for high-volume onboard data processing and pertinent automated information extraction is therefore great.
Exploratory robots are required to send back most of the collected scientific data, unless they become repetitive. The unknown space environment accessible to the sensors is translated into a different, still uninterpreted environment, in the form of computer data banks on Earth. These data banks are then accessible for scientific investigation long after the space mission is over.
The usual purpose of global service robots is to collect time-dependent data in the Earth's environment, whose static properties are well-known. The data are used to determine specific patterns or classes of characteristics and translate these into useful information. For instance, for Landsat and Seasat (Figure 4-4), the data are currently sent to the ground, where they are processed, reduced, annotated, analyzed, and distributed to the user. This process requires up to 3 months for a fully processed satellite image and costs several thousand dollars. The image must then be interpreted by the receiver; i.e., the information must still be extracted by the user.
Projections into the future lead one to speculate on the possibility of highly autonomous exploratory robots in space. Such exploratory robots would communicate to Earth only when contacted or when a significant event occurs and requires immediate attention on Earth. Otherwise, they would collect the data, make appropriate decisions, archive them, and store them onboard. The robots would serve as a data bank, and their computers would be remotely operated by accessing and programming them from Earth whenever the communication link to the robot spacecraft is open. Scientists would be able to interact with the robot by remote terminal. Indeed, the concept of distributed computer systems, presently under investigation in many places, could provide to each instrument its own microcomputer, and scientists could communicate with their respective instruments. They could perform special data processing onboard and request the data to be communi
ated to them in the form desired. Alternatively, they could retrieve particular segments of raw data and perform the required manipulations in their own facilities on Earth.
Prime elements in this link between scientists and distant exploratory robots would be large antenna relay stations in geosynchronous orbit. These stations would also provide data handling and archiving services, especially for inaccessible exploratory robots, e.g., those leaving the solar system.
2. Global Services
Global service robots orbit the Earth. They differ from exploratory robots primarily in the intended application of the collected data. They collect data for public service use on soil conditions, sea states, global crop conditions, weather, geology, disasters, etc. These robots generally acquire and process an immense amount of data. However, only a fraction of the data is of interest to the ultimate user. At the same time, the user often likes to have the information shortly after it has been obtained by the spacecraft. For instance, the value of weather
Figure 4-4. Seasat. The oceanographic satellite's high-data-rate Synthe
tic Aperture Radar imaging device has provided data on ocean waves, coastal regions, and sea ice.
ing image information content is provided, and the user will have to scan catalogs of pictures to find what he or she wants. For such reasons, it is expected that most of the data will not be used again.
The ground operations for Earth orbital missions suffer from problems similar to those of planetary missions. The overall data stream is usually much higher for Earth orbital missions, images are still very costly, and they take up to several months to reach the user.
Present developments in artificial intelligence, machine intelligence, and robotics suggest that, in the future, the groundbased data processing and information extraction functions will be performed onboard the robot spacecraft. Only the useful information would be sent to the ground and distributed to the users, while most of the collected data could be discarded immediately. This would require the robot to be able to decide what data must be retained and how they were to be processed to provide the user with the desired information. For instance, the robot could have a large number of pattern classification templates stored in its memory or introduced by a user with a particular purpose in mind. These templates would represent the characteristics of objects and/or features of interest. The computer would compare the scanned patterns with those stored in its memory. As soon as something of interest appeared, it would examine it with higher resolution, comparing it to a progressively narrower class of templates until recognition had been established to a sufficient degree of confidence. The robot would then contact the appropriate ground station and report its findings and, if required, provide the user with an annotated printout or image. The user would be able to interact with the robot, indeed with his particular instrument, by remote terminal much the same as with a central computer and, depending on intermediate results, modify subsequent processing.
These considerations strongly suggest that technology must be developed so that most ground operation activities can be performed as close as possible to the sensors where the data is collected, namely by the robot in space. However, examining the various ground operations in detail, we conclude that most of those that must remain on the ground could also be automated with advanced machine intelligence techniques. The expected benefits derived from this would be a cost reduction for ground operations of at least an order of magnitude and up to three orders of magnitude for user-ready image information.
3. Utilization of Space Systems
For space exploration and global services, the groundbased mission operations can become extremely complex. A recent example of a planetary exploration mission, and perhaps the most complex to date, is Viking. At times there were several hundred people involved in science data analysis, mission planning, spacecraft monitoring, command sequence generation, data archiving, data distribution, and simulation. Although for earlier space missions sequencing had been determined in advance, on Viking this was done adaptively during the mission. The operational system was designed so that major changes in the mission needed to be defined about 16 days before the spacecraft activity. Minor changes could be made as late as 12 hours before sending a command. The turnaround time of about 16 days and the number of people involved contributes, of course, to sharply increased operational costs. The Viking operations costs (Figure 3-1) are for a 3-month mission. The planned Mars surface rover mission is expected to last 2 years, covering many new sites on the Martian surface. Considering that this mission would be more complex and eight times as long, ground operations would have to be at least ten times as efficient to stay within, or close to, the same relative costs as for Viking.
Space industrialization requires a broader spectrum of robotics and automation capabilities than those identified for space exploration and global services. The multitude of systems and widely varying activities envisioned in space until the end of this century will require the development of space robot and automation technology on a broad scale. It is here that robot and automation technology will have its greatest economic impact. The systems under consideration range
from large antennas and processing and manufacturing stations in Earth orbit to lunar bases, to manned space stations, to satellite power systems of up to 100 km2. These systems are not matched in size by anything on Earth. Their construction and subsequent maintenance will require technologies not yet in use for similar operations on Earth.
Space processing requires a sophisticated technology. First it must be developed and perfected, and then it must be transferred into the commercial arena. Basic types of processes currently envisioned include solidification of melts without convection or sedimentation, processing of molten samples without containers, diffusion in liquids and vapors, and electrophoretic separation of biological substances. It is expected that specialized automated instrumentation will be developed for remote control once the particulars of these processes are worked out and the pressure of commercial requirements becomes noticeable.
During the Viking mission, about 75,000 reels of image data tapes were collected and stored in many separate locations. The images are now identifiable only by the time when and the location where they were taken. No indication regard