With the placement of a space station in low Earth orbit, this Nation will have developed the capability for a permanent human presence in space. To ensure maximum productivity and utility of this space asset, operating personnel must be freed from chores that can be done by machines to devote their time to customer and user needs that cannot be done by machines. Technological advances in a variety of disciplines, collectively called automation and robotics, now permit machines to assume more and more of the burden of system management and operation that has previously been heavily labor-intensive both in the air and on the ground. Future advances, particularly in intelligent robotics, offer the potential of revolutionizing our approach to space operations. Much of the knowledge gained and the advances made will be transferable to terrestrial applications to increase the productivity of the nation's economy as a whole. This paper provides an overview and a summary of the various possible applications and benefits of automation and robotics technology, both near-term and long-term, to the Spabe Station system. Also included are remarks relevant to the National Aeronautics and Space Administration's approach in fostering the development of the needed technologies and views on the benefits to the Nation's economy.

In a complex system such as the manned Space Station, it is deem necessary that many expert systems must perform tasks in a concurrent and cooperative manner. An important question arise is: what cooperative-task-performing models are appropriate for multiple expert systems to jointly perform tasks. The solution to this question will provide a crucial automation design criteria for the Space Station complex systems architecture. Based on a client/server model for performing tasks, we have developed a system that acts as a front-end to support loosely-coupled communications between expert systems running on multiple Symbolics machines. As an example, we use two ART*-based expert systems to demonstrate the concept of parallel symbolic manipulation for power distribution management and dynamic load planner/scheduler in the simulated Space Station environment. This on-going work will also explore other cooperative-task-performing models as alternatives which can evaluate inter and intra expert system communication mechanisms. It will be served as a testbed and a bench-marking tool for other Space Station expert subsystem communication and information exchange.

Many aspects of space station operations involve continuous control of real-time processes. These processes include electrical power system monitoring, propulsion system health and maintenance, environmental and life support systems, space suit checkout, on-board manufacturing, and servicing of attached vehicles such as satellites, shuttles, orbital maneuvering vehicles, orbital transfer vehicles and remote teleoperators. Traditionally, monitoring of these critical real-time processes has been done by trained human experts monitoring telemetry data. However, the long duration of space station missions and the high cost of crew time in space creates a powerful economic incentive for the development of highly autonomous knowledge-based expert control procedures for these space stations. In addition to controlling the normal operations of these processes, the expert systems must also be able to quickly respond to anomalous events, determine their cause and initiate corrective actions in a safe and timely manner. This must be accomplished without excessive diversion of system resources from ongoing control activities and any events beyond the scope of the expert control and diagnosis functions must be recognized and brought to the attention of human operators. Real-time sensor based expert systems (as opposed to off-line, consulting or planning systems receiving data via the keyboard) pose particular problems associated with sensor failures, sensor degradation and data consistency, which must be explicitly handled in an efficient manner. A set of these systems must also be able to work together in a cooperative manner. This paper describes the requirements for real-time expert systems in space station control, and presents prototype implementations of space station expert control procedures in PICON (process intelligent control). PICON is a real-time expert system shell which operates in parallel with distributed data acquisition systems. It incorporates a specialized inference engine with a specialized scheduling portion specifically designed to match the allocation of system resources with the operational requirements of real-time control systems. Innovative knowledge engineering techniques used in PICON to facilitate the development of real-time sensor-based expert systems which use the special features of the inference engine are illustrated in the prototype examples.

In many complex monitoring and diagnostic applications it is useful to build separate expert systems for different subsystems and have them operate (on different computers) mostly in an independent mode with only occasional interactions. There are many advantages of such a scheme including easier knowledge engineering and increased overall system performance. Using an enhanced version of the Lockheed Expert System (LES), we have developed a communicating expert systems for fault diagnosis and fault correction in a prototype for the Space Station Air Revitalization System (ARS). The system consists of three communicating expert systems, one for oxygen generation, one for CO₂ removal and a supervisor for overall control. The three expert system modules communicate via mailboxes. The purpose of this work is to gain an understanding of the problems involved and advantages of using such a communicating expert systems framework.

This paper discusses the requirements for integrating artificial intelligence (AI) techniques with traditional real-time computing in distributed computing environments, and describes an architecture and high-level programming model developed for this purpose. The Multigraph Architecture is a multilayered system, which includes a parallel computation model, the corresponding execution environment and various software tools. The components of the high-level programming model are tailored to the specific properties of the distri-buted computing system, and are centered around the concept of autonomous, communicating objects. The programming model consists of special object types that offer high-level support for the design of complex system components, such as procedural networks, or rule-based systems.

An autonomous mobile robot deals with the empirical world which is never fully predictable, hence it must continually monitor its performance by comparing the actual responses of sensors to their expected responses. Where a discrepancy occurs, the source of the discrepancy must be diagnosed and on-line corrective actions or replanning may be required. The use of a production system for the control of an autonomous robot presents several attractive features: the explicitness and homogeneity of the knowledge representation facilitates explaining, verifying and modifying the rules which determine the robot's behavior; it also permits the incremental extension of the domain of competence. However, real-time operation poses a number of challenges due to the dynamic nature of the data and because the system must frequently deal with a large knowledge base in a limited time. An implementation of a control system is discussed where a large commercial real-time expert system originally designed for industrial process diagnostics was adapted to the control of an autonomous mobile robot for planning, executing and monitoring a set of navigational tasks. One of the essential components of the problem domain is the occurrence of an "unexpected" happening e.g., as new obstacles are moved into the domain during the robot traverse, or when an obstacle undetectable by the long-range sonar sensors is suddenly observed by a proximity sensor. In a recent demonstration of the system, the detection of a problem generated an interrupt alarm, a diagnostic procedure, and a new plan, which was successfully executed in real time.

Artificial intelligence research on automatic planning could result in a new class of management aids to reduce the cost of constructing the Space Station, and would have economically important spinoffs to terrestrial industry as well. Automatic planning programs could be used to plan and schedule launch activities, material deliveries to orbit, construction procedures, and the use of machinery and tools. Numerous automatic planning programs have been written since the 1050's. We describe PARPLAN, a recently-developed experimental automatic planning program written in the AI language Prolog, that can generate plans with parallel activities.

DAISY-DAMP is a prototype of an avant-garde electrical power control system. Although the system is still in the early stages of its development a number of interesting design characteristics are beginning to emerge. Two of the most significant of these features are discussed here. The first has to do with the structure of the system as a set of cooperating agents. The second has to do with the utility, in this domain, of a developmental testbed as a tool for exploring the relevant design space. The utilization of testbeds to explore the potential applicability of recent advances in temporal reasoning and machine learning to subsystem control problems of this sort is also discussed.

The Office of Aeronautics and Space Technology has granted approval for a multi-year sequence of demonstrations to develop, test, and evaluate automation technologies for potential use in the Space Station. This program, called the Systems Autonomy Demonstration Program, is a cooperative venture between NASA Ames Research Center and NASA Johnson Space Center. The Systems Autonomy Demonstration Program has a number of ambitious goals whose attainment is based in part upon results anticipated from a vigorous core research and technology effort.

Aerospace industry interest in autonomy and automation, given fresh impetus by the national goal of establishing a Space Station, is becoming a major item of research and technology development. The promise of new technology arising from research in Artificial Intelligence (AI) has focused much attention on its potential in autonomy and automation. These technologies can improve performance in autonomous control functions that involve planning, scheduling, and fault diagnosis of complex systems. There are, however, many aspects of system and subsystem design in an autonomous system that impact AI applications, but do not directly involve AI technology. Development of a system control architecture, establishment of an operating system within the design, providing command and sensory data collection features appropriate to automated operation, and the use of design analysis tools to support system engineering are specific examples of major design issues. Aspects such as these must also receive attention and technology development support if we are to implement complex autonomous systems within the realistic limitations of mass, power, cost, and available flight-qualified technology that are all-important to a flight project.

This paper discusses some of the implementation issues associated with autonomous management of the Space Station electric energy system. In particular, an executive based structure is proposed to handle the scheduling and real time control of the energy system. The major pieces of this structure are a user interface, the executive itself, a scheduler for the loads and storage, and a real time controller. The least developed of the technologies necessary to realize autonomous management is the scheduler and hence is given primary focus. This paper proposes using the simulated annealing algorithm for the solution of the scheduling problem and presents some preliminary results using this algorithm.

In many applications, such as computer vision, medical imaging, scene forming, and motion animation, surface data is available in the form of two-dimensional (2D) or three-dimensional (3D) contours. Reconstruction of the surface integrates this information into a three-dimensional format. A good surface reconstruction framework must be able to include the surface characteristics inherited in the local structures of the contours. In order to achieve this goal, we have developed a novel surface reconstruction framework which consists of the formulation and solution of three separate sub-problems: (0 contour segmentation; (ii) matching of segments on adjoining contours; (iii) creation of parametric surface patches for joining matched pairs of segments on adjoining contours. New techniques for the solution of the first two subproblems were described in proceeding papers. In the present paper, we discuss a new method for surface patch formation by blending curve segments using spline and sinc blending functions. These blending functions interpolate simultaneously several segments and are chosen such that the boundary conditions are satisfied, and the shape of a segment is propagated along the surface beyond neighboring segments.

Design of NASA's Space Station has begun. During the design cycle, and after activation of the Space Station, the reoccuring need will exist to access not only designs; but also deeper knowledge about the designs, which is only hinted in the design definition. Areas benefiting from this knowledge include training, fault management, and onboard automation. NASA's Artificial Intelligence Office at Johnson Space Center and The MITRE Corporation have conceptualized an approach for capture and storage of design knowledge.

Intelligent control of Space Station will require the coordinated execution of computer programs across a substantial number of computing elements. It will be important to develop large subset of these programs in the form of single programs which execute in a distributed fashion across a number of processors. The single program approach to programming closely coordinated actions of multiple computers allows the advantages of language level software engineering developments, e.g., abstract data types, separate compilation of specifications and implementations, and extensive compile time error checking, to be fully realized across machine boundaries. As yet, however, there are few implementations of distri-buted execution systems. Ada has been adopted for use in the Space Station, and the Ada Language Reference Manual indicates that distributed execution of Ada programs was in the minds of the language designers. However, when considered from the perspective of distributed execution, there are several aspects of the language definition which need further refinement, and a number of difficult trade-off decisions to be made in terms of translation strategies. This paper examines some of the fundamental issues and trade-offs for distributed execution systems for the Ada language.

Mission sequencing consumes a large amount of time and manpower during a space exploration effort. Plan-It is a knowledge-based approach to assist in mission sequencing. Plan-It uses a combined frame and blackboard architecture. This paper reports on the approach implemented by Plan-It and the current applications of Plan-It for sequencing at NASA.

Relative position information concerning an object that is to be acquired, attached to, or manipulated in some way by a robotic system is usually supplied by a known database or through vision information of some kind. Vision systems normally require some degree of intelligence to produce complete position information and therefore are relatively sophisticated, slow, or both. Simple "targets" require some amount of pattern recognition in autonomous operations and do not usually lend themselves to precision applications. This paper describes work on a discrete optical element prototype target which when interrogated by a video camera system, will provide noncontact relative position information about all 6 degrees-of-freedom (DOF). This information is available within the active field of view (FOV) of the transponder and could be processed by microprocessor-based, software algorithms with simple pattern recognition capabilities. The interrogation system (camera) is composed of a standard charge injection device (CID) array video camera, a controllable macrozoom lens, a liquid crystal shutter (LCS), and a point-source multispectral illuminator. This allows the transponder to be used where a standard video camera vision system is needed, or already implemented, and results in a relatively fast system (approximately 10 Hz). A passive optically encoded transponder (POET) implemented in a "stick-on" holographic optical element (HOE) is proposed as a next generation target, to supply relative position information in all 6 DOF for acquisition and precision alignment. In applications requiring maximum bandwidth and resolution, the fact that no "pattern recognition" is required in the proposed system results in the ability to interrogate the transponder in real time with a dedicated nonvision, interrogation system, resulting in a multiorder of magnitude increase in speed. The transponder (target) is configured to provide optimum information for the intended use. Being configurable, it can provide an acquisition signal and alignment information within a wide, operational field of view. The fact that this transponder is a simple, straightforward optical design opens the possibility to implement it in a lightweight, small, economical, "stick-on" holographic alignment target to be used in many industrial and space robotic applications where a classical, standard device is needed.

This paper describes the use of computer graphic simulation techniques to resolve critical design and operational issues for robotic systems used for on-orbit operations. The major design issues in developing effective telerobotic systems are discussed and the use of ROBOSIM, a NASA developed simulation tool, to address these issues is presented. Simulation plans for the Space Station and the Orbital Maneuvering Vehicle are presented and discussed.

The multi-billion dollar Space Station that will orbit the skies in the early 1990's will be the most complex system ever placed in orbit by man. The station will provide a multitude of services, from collecting data from distant stars to repairing failed satellites and manufacturing extremely pure pharmacological products.

Most robots we are familiar with are general purpose robots, designed by their manufacturers to perform a wide variety of tasks in a number of operating conditions. In designing these robots, the manufacturers usually attempt to maximize parameters such as payload and velocity while minimizing manufacturing cost. For space applications, there are other design criteria which must be met. For instance, unlike ground based robotic systems, for which an "unlimited" amount of power is available; space based robotic systems have only a limited amount of available power. In addition, the use of robotics in space in the immediate future will be primarily for one-of-a kind applications, therefore designing an arm to minimize manufacturing cost is a low priority. To allow for the successful utilization of robots in space, new techniques are needed to meet these design criteria. In this paper, the author will present a method to give robot designers a starting point for designing "optimal" robots for space applications.

An identified goal for initial Space Station's operations is to have a telerobotic system that can perform assembly and maintenance tasks on the station. Servicing and assembly which are candidates for automation include tasks which are potentially hazardous to humans and tasks which are repetitious. This paper will describe the architecture of an integrated telerobotics laboratory which is being used for research on the mechanisms, controls, sensing, and operator interface required to accomplish space telerobotic tasks. The Intelligent Systems Research Laboratory (ISRL) uses a hierarchical structure of functionally distributed computers communicating over both parallel and high-speed serial data paths in conjunction with a modular system simulation program to conduct studies of advanced telerobotic systems. Multiple processes perform motion planning, operator communications, forward and inverse kinematics, control/sensor fusion, and I/O processing while communicating through common memory on a VAX host computer. Additional hardware elements of the simulation include a symbolic processor, a high-speed computer graphics system, manipulators, and a vision processor. Two manipulators can be operated under teleoperator control or can be supervised by the operator while performing a sequence of elementary operations using force, torque, and vision sensing. This paper describes the architecture and capability of the laboratory and discusses recent telerobotic studies related to satellite servicing and space assembly.

Based on generic operational and computational requirements associated with the control of telerobots in Earth orbit, a multibus-based distributed but integrated computing architecture is proposed. An experimental system of that kind under development at the Jet Propulsion Laboratory (JPL) is briefly described. It uses Intel Multibus I at both control station and remote robot (telerobot) computing nodes. NS32000 series CPU boards constitute the processing elements of each node. An essential element within each multibus is a Unified (or Universal) Computer Control Subsystem (UCCS) for telerobot and control station motor components. The two multibus-based computing nodes can be linked by parallel or high speed serial links for real-time data transmission and for closing the real-time bilateral (force-reflecting) control loop between telerobot and control station. The experimental system is briefly commented, followed by a brief discussion of future development plans and possibilities.

We report the ongoing development of a telerobot technology demonstrator. The demonstrator is implemented as a laboratory-based research testbed, and will show proof-of-concept for supervised automation of space assembly, servicing, and repair operations. The demonstrator system features a hierarchically layered intelligent control architecture which enables automated planning and run-time sequencing of complex tasks by a supervisory human operator. The demonstrator also provides a full bilateral force-reflecting hand control teleoperations capability. The operator may switch smoothly between the automated and teleoperated tasking modes in run-time, either on a preplanned or operator-designated basis.

The on-orbit repair sequence of the Solar Maximum Mission (SMM) in April 1984 is described, and lessons derived from this experience are noted. Implications for serviceability in the design of future spacecraft and instruments are illustrated by the case of the Hubble Space Telescope (HST) and its second generation spectrograph.

The Space Station will be a very complex, closed system. A relatively small crew will be required to manage the day-to-day activities of the Station and to tend the experimental and commercial payloads. Decisions will be required concerning the operation and maintenance of all the elements of the Station. The emerging technology of artificial intelligence is being utilized to build expert system software which will provide the routine reasoning and problem-solving capabilities required to off-load the crew for more productive tasks. The complex problem of building these expert systems and the man-machine interfaces which make them effective are the subjects of this paper.

Space-based crews are expected to interact with highly automated and possibly intelligent systems and to perform these interactions with often little or no prior training, or on an infrequent or sporadic basis. These activities will characterize a new role for space-based crews, that of supervisory control. Supervisory control tasks in turn define a new set of requirements for Space Station man-machine interface (MMI) design: (1) multi-function display and control hardware, (2) displays that enhance the crew person's "mental model" of invisible processes, (3) highly supportive man-machine dialogue, including special features to support dialogue with expert systems, (4) incorporation of machine intelligence into the MMI itself to provide a seemingly uniform interface to numerous processes, data bases, and expert systems, and (5) electronic documentation. A discussion of these concepts is illustrated by examples from recent MMI designs, including a multi-function display and control system developed for the Space Shuttle, an MMI system developed for NASA JSC for the Space Station environmental control and life support system, ATOZ--an intelligent interface system, and VIMAD--an electronic documentation system for maintenance procedures.

The domains most amenable to techniques based on artificial intelligence (AI) are those that are systematic or for which a systematic domain can be generated. In aerospace systems, many operational tasks are sys-tematic owing to the highly procedural nature of the applications. However, aerospace applications can also be nonprocedural, particularly in the event of a failure or an unexpected event. Several techniques are discussed for designing automated systems for real-time, dynamic environments, particularly when a "breakdown" occurs. A breakdown is defined as operation of an automated system outside its predetermined, conceptual domain.

The mission lifetime of future spacecraft will be greatly extended by the routine performance of repair and resupply operations. There will be significant cost and safety advantages if these operations can be performed with minimum human involvement. A key step in any such operation will be rendezvous and docking. Each docking operation requires knowledge of the three-dimensional position and orientation (attitude) of the associated spacecraft. In this paper, an approach for solving the attitude determination problem for satellites is presented. This technique uses standard CCD video and requires the identification of the camera plane coordinates of at least three known points on the object. To facilitate detection of these points, retro-reflectors are mounted in a known configuration on the satellite's surface and are illuminated by a spotlight adjacent to the camera. A ring of black nonreflective material is placed around the border of each reflector to aid the image processing. The reflectors are separated from the background using features such as brightness, shape, and color. Besides distinguishing the reflectors from the background it is also necessary to distinguish them from each other. To do this we arranged the identical disk reflectors in a specific pattern, which allows the correspondence to be determined from the image locations alone. The use of identical disks has enabled development of robust algorithms that can reliably identify reflectors in the presence of adverse lighting conditions, such as secondary reflections off thermal blankets, or the earth or sun in the background.

In any computer vision system capable of monitoring sufficiently-complex, dynamic scenes, the following two properties will be essential in order to ensure efficient and timely operation: the overall system is designed such that the processes concentrate their computational effort in areas where important features are likely to be identified, not wasting much time on irrelevant data, and the analysis of the image data can be broken down and handled in parallel by cooperating processes. In this paper a computerized sys-tem for visually monitoring life sciences experiments on board a space station will be described. In addition, a hypothetical monitoring task will be discussed in order to demonstrate how knowledge of the moni-tored environment can aid in both the division of labor among parallel, coordinated, image-analyzing processes, and also the ability of the system to focus its efforts on relevant information. The approach taken here is to conceive of the vision system as a particular type of expert system in which knowledge of the monitored domain is incorporated in the form of rules.

An experimental telerobotics (TR) simulation is described suitable for studying human operator (H.O.) performance. Simple manipulator pick-and-place and tracking tasks allowed quantitative comparison of a number of calligraphic display viewing conditions. The Ames-Berkeley enhanced perspective display was utilized in conjunction with an experimental helmet mounted display system (HM0IFIt provided stereoscopic enhanced views. Two degree-of-freedom rotations of the head were measured with a Helmholtz coil instrument and these angles used to compute a directional conical window into a 3-D simulation. The vector elements within the window were then transformed by projective geometry calculations to an intermediate stereoscopic display, received by two video cameras and imaged onto the HID mini-display units (one-inch CRT video receivers) mounted on the helmet. An introduced communication delay was found to oroduce decrease in performance. In considerable part, this difficulty could be compensated for by preview control information. That neurological control of normal human movement contains a sampled data period of 0.2 seconds may relate to this robustness of H.0. control to delay. A number of control modes could be compared in this TR simulation, including displacement, rate iiracceleratory control using position and force joysticks. A homeomorphic controller turned out to be no better than joysticks; the adaptive properties of the H.O. can apparently permit quite good control over a variety of controller configurations and control modes. Training by optimal control example seemed helpful in preliminary experiments.

Current expert system technology does not permit complete automatic fault diagnosis; significant levels of human intervention are still required. This requirement dictates a need for a division of labor that recognizes the strengths and weaknesses of both human and machine diagnostic skills. Relevant findings from the literature on human cognition are combined with the results of reviews of aircrew performance with highly automated systems to suggest how the interface of a fault diagnostic expert system can be designed to assist human operators in verifying machine diagnoses and guiding interactive fault diagnosis. It is argued that the needs of the human operator should play an important role in the design of the knowledge base.

In this paper attempt is made to improve the overall performance of a stripmap image sensor through the implementation with AI. The virtual range of scanning thus obtained could be two to three times as large as the original range. The approach suggested is mainly a predictive control with which the viewing direction of the sensor is intermittently centered at the centroid of the target information acquired during the preceding scans so that maximum amount of information on the successive target segments can be obtained. Simulation experiment conducted verifies the feasibility of the suggested approach.