ALBERT

Description: The uncertainties inherent in the dynamics of flexible spacecraft make robustness questions very important when designing vibration suppression systems for these vehicles. This often leads to the use of sensors and actuators which are collocated on the structure, so as to avoid the potentially destabilizing problem of unknown modal phase differences between noncollocated sensors and actuators. The closed-loop performance attainable is, of course, reduced if only collocated pairs are considered. The object of the present study is to investigate whether the special properties of the transmission zeros of such structures can be used to quantify these performance limitations. The approach is based on recent results concerning the transmission zeros of flexible structures.

Description: A generalized master/slave technique and experimental results for coordinated control of two arms rigidly grasping an object is described. An interactive program has been developed to allow a user the flexibility to select appropriate control modes for a given experiment. This interface allows for control gain adjustments. The results of several experiments performed on this system to demonstrate its capabilities such as transporting an object with or without induced internal forces and movement of a constrained object are offered. The system is further developed to achieve a so-called shared control mode in which an operator specifies the free motion trajectory for a point on the object of manipulation via a joystick while the autonomous control system is used for coordination and control of the arms.

Description: Algorithmic complexity is discussed as a computational counterpart to the second law of thermodynamics. It is shown that algorithmic complexity, which is a measure of randomness, sets limits on the thermodynamic cost of computations and casts a new light on the limitations of Maxwell's demon. Algorithmic complexity can also be used to define distance between binary strings.

Description: Two procedures for the evaluation of the performance of expert systems are illustrated: one procedure evaluates predictive accuracy; the other procedure is complementary in that it uncovers the factors that contribute to predictive accuracy. Using these procedures, it is argued that expert systems should be more accurate than human experts in two senses. One sense is that expert systems must be more accurate to be cost-effective. Previous research is reviewed and original results are presented which show that simple statistical models typically perform better than human experts for the task of combining evidence from a given set of information sources. The results also suggest the second sense in which expert systems should be more accurate than human experts. They reveal that expert systems should share factors that contribute to human accuracy, but not factors that detract from human accuracy. Thus the thesis is that one should both require and expect systems to be more accurate than humans.

Description: The decoder error probability for Reed-Solomon codes (more generally, linear maximum distance separable codes) is examined. McEliece and Swanson offered an upper bound on P sub E (u), the decoder error probability given that u symbol errors occur. This upper bound is slightly greater than Q, the probability that a completely random error pattern will cause decoder error. By using a combinatoric technique, the principle of inclusion and exclusion, an exact formula for P sub E (u) is derived. The P sub E (u) for the (255,223) Reed-Solomon Code used by NASA, and for the (31,15) Reed-Solomon code (JTIDS code), are calculated using the exact formula, and the P sub E (u)'s are observed to approach the Q's of the codes rapidly as u gets larger. An upper bound for the expression is derived, and is shown to decrease nearly exponentially as u increases. This proves analytically that P sub E (u) indeed approaches Q as u becomes large, and some laws of large numbers come into play.

Description: Simple state-space formulas are derived for all controllers solving the standard H(infinity) problem of finding, for a given number gamma greater than 0, all controllers such that the H(infinity) norm of the closed-loop transfer function is (strictly) less than gamma. It is known that a controller exists if and only if the unique stabilizing solutions to two algebraic Riccati equations are positive definite and the spectral radius of their product is less than gamma squared. Under these conditions, a parameterization of all controllers solving the problem is given as a linear fractional transformation (LFT) on a contractive stable free parameter. The state dimension of the coefficient matrix for the LFT, constructed using the two Riccati solutions, equals that of the plant and has a separation structure reminiscent of classical LQG (i.e., H2) theory. A standard H2 solution is developed in parallel.

Description: Recursive forward dynamics algorithms are developed and presented for an arbitrary number of robot arms moving a commonly held object. The multiarm forward dynamics problem is to find the angular accelerations at the joints and the contact forces that the arms impart to the task object. The problem also involves finding the acceleration of this object. The multiarm forward dynamics solutions provide a thorough physical and mathematical understanding of the way several arms behave in response to a set of applied joint moments. Such an understanding simplifies and guides the subsequent control design and experimentation process. The forward dynamics algorithms also provide the necessary analytical foundation for conducting analysis and simulation studies. The multiarm algorithms are based on the filtering and smoothing approach recently advanced for single-arm dynamics, and they can be built up modularly from the single-arm algorithms. The algorithms compute recursively the joint angle accelerations, the contact forces, and the task-object accelerations. Algorithms are also developed to evaluate in closed form the linear transformations from the active joint moments to the joint angle accelerations, to the task object accelerations, and to the task-object contact forces. A possible computing architecture is presented as a precursor to a more complete investigation of the computational performance of the dynamics algorithms.

Description: A status review is presented for a generic blackboard-based distributed problem-solving environment in which multiple-agent cooperation can be effected. This environment is organized into a shared information panel, a chairman control panel, and a metaplanning panel. Each panel contains a number of embedded AI techniques that facilitate its operation and that provide heuristics for solving the underlying team-agent decision problem. The status of these panels and heuristics is described along with a number of robustness considerations. The techniques for each of the three panels and for four sets of paradigm-related advances are described, along with selected results from classroom teaching experiments and from three applications.

Description: The linear optimal control problem of an nth-order time-invariant dynamic system with a quadratic performance functional is usually solved by the Hamilton-Jacobi approach. This leads to the solution of the differential matrix Riccati equation with a terminal condition. The bulk of the computation for the optimal control problem is related to the solution of this equation. There are various algorithms in the literature for solving the matrix Riccati equation. However, computational complexities and storage requirements as a function of numbers of state variables, control variables, and sensors are not available for all these algorithms. In this work, the computational complexities and storage requirements for some of these algorithms are given. These expressions show the immensity of the computational requirements of the algorithms in solving the Riccati equation for large-order systems such as the control of highly flexible space structures. The expressions are also needed to compute the speedup and efficiency of any implementation of these algorithms on concurrent machines.

Description: The authors propose a model, or filter, for debiasing opinions from multiple experts and combining them into a single consistent estimate of some variable of interest. A distinguishing feature of the approach consists of making the calibration of experts an integral part of filtering. This enables the filter to learn from previous experience with the experts. The theoretical development takes a Bayesian perspective, using de Finetti's (1964) notion of exchangeability. Experimental results with a preliminary computer implementation of the filter show that its estimates are better than those from comparable filters that do not involve calibration.

Description: The approximate inversion of an internally unknown linear system, given by its impulse response sequence, by an inverse system having a finite impulse response, is considered. The recursive least-squares procedure is shown to have an exact initialization, based on the triangular Toeplitz structure of the matrix involved. The proposed approach also suggests solutions to the problem of system identification and compensation.

Description: A methodology for robust eigenstructure assignment for multivariable feedback systems is presented. The algorithm is based upon a pole placement technique using projections onto subspaces of admissible eigenvectors. New ideas are introduced to generate target (desired) sets of unitary eigenvectors and determine optimal feasible eigenvectors in a least-square sense. Useful connections are established between the pole-placement by independent modal space control and the method introduced in this paper. A multicriterion optimization algorithm is also presented, which takes efficient advantage of the present eigenstructure assignment method. These developments show significant improvement over an earlier version of this algorithm in both computational cost and accuracy. This optimization process appears to be numerically robust and suitable for high-dimensional multicriterion optimizations; it is especially attractive for computer-aided design of control systems.

Description: This paper corrects an unclear statement of the conventional Kalman filter implementation as presented by M. H. Verhaegen and P. van Dooren in Numerical aspects of different Kalman filter implementations, IEEE Trans. Automat. Contr., v. AC-31, no. 10, pp. 907-917, 1986. It is shown that the habitual, incorrect implementation of the Kalman filter has been the major cause of its insensitivity to the so-called loss-of-symmetry phenomenon.

Description: Problems of implementing rule-based expert systems using fuzzy sets are considered. A fuzzy logic software development shell is used that allows inclusion of both crisp and fuzzy rules indecision making and process control problems. Results are given that compare this type of expert system to a human expert in some specific applications. Advantages and disadvantages of such systems are discussed.

Description: A large scale domain-independent spatial data management expert system that serves as a front-end to databases containing spatial data is described. This system is unique for two reasons. First, it uses spatial search techniques to generate a list of all the primary keys that fall within a user's spatial constraints prior to invoking the database management system, thus substantially decreasing the amount of time required to answer a user's query. Second, a domain-independent query expert system uses a domain-specific rule base to preprocess the user's English query, effectively mapping a broad class of queries into a smaller subset that can be handled by a commercial natural language processing system. The methods used by the spatial search module and the query expert system are explained, and the system architecture for the spatial data management expert system is described. The system is applied to data from the International Ultraviolet Explorer (IUE) satellite, and results are given.

Description: Existing NASA supported scientific data bases are usually developed, managed and populated in a tedious, error prone and self-limiting way in terms of what can be described in a relational Data Base Management System (DBMS). The next generation Earth remote sensing platforms, i.e., Earth Observation System, (EOS), will be capable of generating data at a rate of over 300 Mbs per second from a suite of instruments designed for different applications. What is needed is an innovative approach that creates object-oriented databases that segment, characterize, catalog and are manageable in a domain-specific context and whose contents are available interactively and in near-real-time to the user community. Described here is work in progress that utilizes an artificial neural net approach to characterize satellite imagery of undefined objects into high-level data objects. The characterized data is then dynamically allocated to an object-oriented data base where it can be reviewed and assessed by a user. The definition, development, and evolution of the overall data system model are steps in the creation of an application-driven knowledge-based scientific information system.

Description: This paper describes a novel data acquisition system for use with wind-tunnel probe-based measurements, which incorporates a degree of specific fluid dynamics knowledge into a simple expert system-like control program. The concept was developed with a rudimentary expert system coupled to a probe positioning mechanism operating in a small-scale research wind tunnel. The software consisted of two basic elements, a general-purpose data acquisition system and the rulebased control element to take and analyze data and supplying decisions as to where to measure, how many data points to take, and when to stop. The system was validated in an experiment involving a vortical flow field, showing that it was possible to increase the resolution of the experiment or, alternatively, reduce the total number of data points required, to achieve parity with the results of most conventional data acquisition approaches.

Description: The spatially constrained clustering (SCC) iterative parallel region-growing technique is applied to image analysis. The SCC algorithm is implemented on the massively parallel processor at NASA Goddard. Most previous region-growing approaches have the drawback that the segmentation produced depends on the order in which portions of the image are processed. The ideal solution to this problem (merging only the single most similar pair of spatially adjacent regions in the image in each iteration) becomes impractical except for very small images, even on a massively parallel computer. The SCC algorithm overcomes these problems by performing, in parallel, the best merge within each of a set of local, possibly overlapping, subimages. A region-splitting stage is also incorporated into the algorithm, but experiments show that region splitting generally does not improve segmentation results. The SCC algorithm has been tested on various imagery data, and test results for a Landsat TM image are summarized.

Description: A suite of telerobotic tasks has been compiled and assessed for the purpose of selecting viable tasks for near and far term laboratory demonstrations. The primary intent of developing the task is to provide some technical guidelines, with supporting data, for focusing laboratory demonstrations toward application domains that address a wide array of potential telerobot tasks and required technologies. This wide application would then result in a rich technology development environment to meet the broad task requirements of a system such as the Flight Telerobot Servicer. The methodology and results of the telerobot task assessment are described, including a ranking of the final select suite of major tasks. The presented along with guidelines for both interpreting the task ranking results and setting programmatic objectives based on these results. Detailed data about the task candidates and their respective levels of complexity, task primitive actions, and the actual relative measures of task worth as associated with key tradeoff variables such as cost, available research resources, technology availability, and importance to the user community are also presented.

Description: The results of using a new technique for shaping inputs to a model of the space shuttle Remote Manipulator System (RMS) are presented. The shapes inputs move the system to the same location that was originally commanded, however, the oscillations of the machine are considerably reduced. An overview of the new shaping method is presented. A description of RMS model is provided. The problem of slow joint servo rates on the RMS is accommodated with an extension of the shaping method. The results and sample data are also presented for both joint and three-dimensional cartesian motions. The results demonstrate that the new shaping method performs well on large, telerobotic systems which exhibit significant structural vibration. The new method is shown to also result in considerable energy savings during operations of the RMS manipulator.

Description: Four separate projects recently completed or in progress at the MIT Man-Machine Systems Laboratory are summarized. They are: a decision aid for retrieving a tumbling satellite in space; kinematic control and graphic display of redundant teleoperators; real time terrain/object generation: a quad-tree approach; and two dimensional control for three dimensional obstacle avoidance.

Description: The implementation and experimental validation of a new direct adaptive control scheme on a PUMA 560 industrial robot is described. The testbed facility consists of a Unimation PUMA 560 six-jointed robot and controller, and a DEC MicroVAX II computer which hosts the Robot Control C Library software. The control algorithm is implemented on the MicroVAX which acts as a digital controller for the PUMA robot, and the Unimation controller is effectively bypassed and used merely as an I/O device to interface the MicroVAX to the joint motors. The control algorithm for each robot joint consists of an auxiliary signal generated by a constant-gain Proportional plus Integral plus Derivative (PID) controller, and an adaptive position-velocity (PD) feedback controller with adjustable gains. The adaptive independent joint controllers compensate for the inter-joint couplings and achieve accurate trajectory tracking without the need for the complex dynamic model and parameter values of the robot. Extensive experimental results on PUMA joint control are presented to confirm the feasibility of the proposed scheme, in spite of strong interactions between joint motions. Experimental results validate the capabilities of the proposed control scheme. The control scheme is extremely simple and computationally very fast for concurrent processing with high sampling rates.

Description: A recursive algorithm is developed for the solution of the simulation dynamics problem for a chain of rigid bodies. Arbitrary joint constraints are permitted, that is, joints may allow translational and/or rotational degrees of freedom. The recursive procedure is shown to be identical to that encountered in a discrete-time optimal control problem. For each relevant quantity in the multibody dynamics problem, there exists an analog in the context of optimal control. The performance index that is minimized in the control problem is identified as Gibbs' function for the chain of bodies.

Description: Space telerobots are recognized to require cooperation with human operators in various ways. Multi-level manual and autonomous control superposition in telerobot task execution is described. The object model, the structured master-slave manipulation system, and the motion understanding system are proposed to realize the concept. The object model offers interfaces for task level and object level human intervention. The structured master-slave manipulation system offers interfaces for motion level human intervention. The motion understanding system maintains the consistency of the knowledge through all the levels which supports the robot autonomy while accepting the human intervention. The superposing execution of the teleoperational task at multi-levels realizes intuitive and robust task execution for wide variety of objects and in changeful environment. The performance of several examples of operating chemical apparatuses is shown.

Description: A classical control perspective is used to characterize performance constraints and evaluate compensation techniques for teleoperation with delay. Use of control concepts such as open and closed loop performance, stability, and bandwidth yield insight to the delay problem. Teleoperator performance constraints are viewed as an open loop time delay lag and as a delay-induced closed loop bandwidth constraint. These constraints are illustrated with a simple analytical tracking example which is corroborated by a real time, 'man-in-the-loop' tracking experiment. The experiment also provides insight to those controller characteristics which are unique to a human operator. Predictive displays and feedforward commands are shown to provide open loop compensation for delay lag. Low pass filtering of telemetry or feedback signals is interpreted as closed loop compensation used to maintain a sufficiently low bandwidth for stability. A new closed loop compensation approach is proposed that uses a reactive (or force feedback) hand controller to restrict system bandwidth by impeding operator inputs.

Description: Hierarchical structures offer numerous advantages over conventional structures for the control of telerobotic systems. A hierarchically organized system can be controlled via undetailed task assignments and can easily adapt to changing circumstances. The distributed and modular structure of these systems also enables fast response needed in most telerobotic applications. On the other hand, most of the hierarchical structures proposed in the literature are based on functional properties of a system. These structures work best for a few given functions of a large class of systems. In telerobotic applications, all functions of a single system needed to be explored. This approach requires a hierarchical organization based on physical properties of a system and such a hierarchical organization is introduced. The decomposition, organization, and control of the hierarchical structure are considered, and a system with two robot arms and a camera is presented.

Description: Kolmogorov's axiomatic principles of the probability theory, are reconsidered in the scope of their applicability to the processes of knowledge acquisition and interpretation. The model of uncertainty generation is modified in order to reflect the reality of engineering problems, particularly in the area of intelligent control. This model implies algorithms of learning which are organized in three groups which reflect the degree of conceptualization of the knowledge the system is dealing with. It is essential that these algorithms are motivated by and consistent with the multiresolutional model of knowledge representation which is reflected in the structure of models and the algorithms of learning.

Description: A neural network for controlling the configuration of frame structure with elastic members is proposed. In the present network, the structure is modeled not by using the relative angles of the members but by using the distances between the joint locations alone. The relationship between the environment and the joints is also defined by their mutual distances. The analog neural network attains the reaching motion of the manipulator as a minimization problem of the energy constructed by the distances between the joints, the target, and the obstacles. The network can generate not only the final but also the transient configurations and the trajectory. This framework with flexibility and parallelism is very suitable for controlling the Space Telerobotic systems with many degrees of freedom.

Description: Two neural learning controller designs for manipulators are considered. The first design is based on a neural inverse-dynamics system. The second is the combination of the first one with a neural adaptive state feedback system. Both types of controllers enable the manipulator to perform any given task very well after a period of training and to do other untrained tasks satisfactorily. The second design also enables the manipulator to compensate for unpredictable perturbations.

Description: When plans are used as programs for controlling the action of autonomous or teleoperated robots, their abstract representation can easily obscure a great deal of the critical knowledge that originally led to the planned course of action. An autonomous vehicle experiment is highlighted which illustrates how the information barriers created by abstraction can result in undesirable action. It is then shown how the same task can be performed correctly using plans as a resource for action. As a result of this simple change in outlook, problems requiring opportunistic reaction to unexpected changes in the environment can be solved.

Description: Researchers describe a newly-developed artificial neural network algorithm for solving constraint satisfaction problems (CSPs) which includes a learning component that can significantly improve the performance of the network from run to run. The network, referred to as the Guarded Discrete Stochastic (GDS) network, is based on the discrete Hopfield network but differs from it primarily in that auxiliary networks (guards) are asymmetrically coupled to the main network to enforce certain types of constraints. Although the presence of asymmetric connections implies that the network may not converge, it was found that, for certain classes of problems, the network often quickly converges to find satisfactory solutions when they exist. The network can run efficiently on serial machines and can find solutions to very large problems (e.g., N-queens for N as large as 1024). One advantage of the network architecture is that network connection strengths need not be instantiated when the network is established: they are needed only when a participating neural element transitions from off to on. They have exploited this feature to devise a learning algorithm, based on consistency techniques for discrete CSPs, that updates the network biases and connection strengths and thus improves the network performance.

Description: A generalized optimality criterion method consisting of a dual problem solver combined with a compound scaling algorithm was implemented in the multidisciplinary design tool, ASTROS. This method enables, for the first time in a production design tool, the determination of a minimum weight design using thousands of independent structural design variables while simultaneously considering constraints on response quantities in several disciplines. Even for moderately large examples, the computational efficiency is improved significantly relative to the conventional approach.

Description: An approach for combined control-structure optimization keyed to enhancing early design trade-offs is outlined and illustrated by numerical examples. The approach employs a homotopic strategy and appears to be effective for generating families of designs that can be used in these early trade studies. Analytical results were obtained for classes of structure/control objectives with linear quadratic Gaussian (LQG) and linear quadratic regulator (LQR) costs. For these, researchers demonstrated that global optima can be computed for small values of the homotopy parameter. Conditions for local optima along the homotopy path were also given. Details of two numerical examples employing the LQR control cost were given showing variations of the optimal design variables along the homotopy path. The results of the second example suggest that introducing a second homotopy parameter relating the two parts of the control index in the LQG/LQR formulation might serve to enlarge the family of Pareto optima, but its effect on modifying the optimal structural shapes may be analogous to the original parameter lambda.

Description: Viewgraphs describing the design and testing of an intelligent decision support system called OFMspert are presented. In this expert system, knowledge about the human operator is represented through an operator/system model referred to as the OFM (Operator Function Model). OFMspert uses the blackboard model of problem solving to maintain a dynamic representation of operator goals, plans, tasks, and actions given previous operator actions and current system state. Results of an experiment to assess OFMspert's intent inferencing capability are outlined. Finally, the overall design philosophy for an intelligent tutoring system (OFMTutor) for operators of complex dynamic systems is summarized.

Description: The design, implementation, and evaluation of an Operator Function Model intelligent tutoring system (OFMTutor) is presented. OFMTutor is intended to provide intelligent tutoring in the context of complex dynamic systems for which an operator function model (OFM) can be constructed. The human operator's role in such complex, dynamic, and highly automated systems is that of a supervisory controller whose primary responsibilities are routine monitoring and fine-tuning of system parameters and occasional compensation for system abnormalities. The automated systems must support the human operator. One potentially useful form of support is the use of intelligent tutoring systems to teach the operator about the system and how to function within that system. Previous research on intelligent tutoring systems (ITS) is considered. The proposed design for OFMTutor is presented, and an experimental evaluation is described.

Description: A strategy to evaluate an expert system is formulated. The strategy proposed is based on finding an equivalent classifier to an expert system and evaluate that classifier with respect to an optimal classifier, a Bayes classifier. Here it is shown that for the rules considered an equivalent classifier exists. Also, a brief consideration of meta and meta-meta rules is included. Also, a taxonomy of expert systems is presented and an assertion made that an equivalent classifier exists for each type of expert system in the taxonomy with associated sets of underlying assumptions.

Description: Adaptive/general learning algorithms using varying neural network models are considered for the intelligent control of robotic arm plus dextrous hand/manipulator systems. Results are summarized and discussed for the use of the Barto/Sutton/Anderson neuronlike, unsupervised learning controller as applied to the stabilization of an inverted pendulum on a cart system. Recommendations are made for the application of the controller and a kinematic analysis for trajectory planning to simple object retrieval (chase/approach and capture/grasp) scenarios in two dimensions.

Description: Some recent problems are no longer formulated in terms of imprecise facts, missing data or inadequate measuring devices. Instead, questions pertaining to knowledge and information itself arise and can be phrased independently of any particular area of knowledge. The problem considered in the present work is how to model a problem solver that is trying to find the answer to some query. The problem solver has access to a large number of knowledge systems that specialize in diverse features. In this context, feature means an indicator of what the possibilities for the answer are. The knowledge systems should not be accessed more than once, in order to have truly independent sources of information. Moreover, these systems are allowed to run in parallel. Since access might be expensive, it is necessary to construct a management policy for accessing these knowledge systems. To help in the access policy, some control knowledge systems are available. Control knowledge systems have knowledge about the performance parameters status of the knowledge systems. In order to carry out the double goal of estimating what units to access and to answer the given query, diverse pieces of evidence must be fused. The Dempster-Shafer Theory of Evidence is used to pool the knowledge bases.

Description: A preliminary study of SOAR, a general intelligent architecture for automated problem solving and learning, is presented. The underlying principles of universal subgoaling and chunking were applied to a simple, yet representative, problem in artificial intelligence. A number of problem space representations were examined and compared. It is concluded that learning is an inherent and beneficial aspect of problem solving. Additional studies are suggested in domains relevant to mission planning and to SOAR itself.

Description: An overview of the research in the field of intelligent tutorial systems (ITS) is provided. The various approaches in the design and implementation of ITS are examined and discussed in the context of problem solving in an environment of a complex dynamic system (CDS). Issues pertaining to a CDS and the nature of human problem solving especially in light of a CDS are considered. An overview of the architecture of an ITS is provided as the basis for the in-depth examination of various systems. Finally, the implications for the design and evaluation of an ITS are discussed.

Description: Viewgraphs providing an overview of activities concerned with the development and testing of the Operator Function Model (OFM) expert system (OFMspert) are presented. The OFM is a mathematical tool for representing operator interaction with predominantly automated space ground control systems. OFM provides cognitive task analysis and served as the basis for the design of an intelligent operator's associate called OMFspert. An experimental implementation of OFMspert, referred to as Ally, was developed. An empirical evaluation of Ally was conducted to determine the effectiveness of a supervisory control team consisting of a human operator and Ally versus a control team consisting of two human operators. The experiment was carried out in the GT-MSOCC (Georgia Tech MultiSatellite Operations Control Center) domain, a research tool consisting of a high fidelity implementation of the operator interface to a GSFC ground control system. The viewgraphs outline the experimental design, operator performance measures, and preliminary results.

Description: Computations required for temporal reasoning can be prohibitively expensive if fully general representations are used. Overly simple representations, such as totally ordered sequence of time points, are inadequate for use in a nonlinear task planning system. A middle ground is identified which is general enough to support a capable nonlinear task planner, but specialized enough that the system can support online task planning in real time. A Temporal Logic System (TLS) was developed during the Intelligent Task Automation (ITA) project to support robotic task planning. TLS is also used within the ITA system to support plan execution, monitoring, and exception handling.

Description: The in-orbit operations, like space structures inspection, servicing and repairing, is expected to be one of the most significant technological area for application and development of Robotics and Automation in Space Station environment. The Italian National Space Plan (PSN) has started up its strategic programme SPIDER (Space Inspection Device for Extravehicular Repairs), which is scheduled in three phases, with the final goal of performing docking and precision repairing in the Space Station environment. SPIDER system is an autonomous integrated space robot, using mature Artificial Intelligence tools and technics for its operational control. The preliminary results of a study on the information architecture of the spacecraft are described.

Description: Some solutions to crucial issues regarding the competent performance of an autonomously operating robot are identified; namely, that of handling multiple and variable data sources containing overlapping information and maintaining coherent operation while responding adequately to changes in the environment. Support for the ideas developed for the construction of such behavior are extracted from speculations in the study of cognitive psychology, an understanding of the behavior of controlled mechanisms, and the development of behavior-based robots in a few robot research laboratories. The validity of these ideas is supported by some simple simulation experiments in the field of mobile robot navigation and guidance.

Description: A concept of nested hierarchical (multi-resolutional, pyramidal) information (knowledge) processing is introduced for a variety of systems including data and/or knowledge bases, vision, control, and manufacturing systems, industrial automated robots, and (self-programmed) autonomous intelligent machines. A set of practical recommendations is presented using a case study of a multiresolutional object representation. It is demonstrated here that any intelligent module transforms (sometimes, irreversibly) the knowledge it deals with, and this tranformation affects the subsequent computation processes, e.g., those of decision and control. Several types of knowledge transformation are reviewed. Definite conditions are analyzed, satisfaction of which is required for organization and processing of redundant information (knowledge) in the multi-resolutional systems. Providing a definite degree of redundancy is one of these conditions.

Description: The planning of dexterous grasps for multifingered robot hands operating in uncertain environments is covered. A sensor-based approach to the planning of a reach path prior to grasping is first described. An on-line, joint space finger path planning algorithm for the enclose phase of grasping was then developed. The algorithm minimizes the impact momentum of the hand. It uses a Preshape Jacobian matrix to map task-level hand preshape requirements into kinematic constraints. A master slave scheme avoids inter-finger collisions and reduces the dimensionality of the planning problem.

Description: An efficient, adaptive neural learning paradigm for addressing the inverse kinematics of redundant manipulators is presented. The proposed methodology exploits the infinite local stability of terminal attractors - a new class of mathematical constructs which provide unique information processing capabilities to artificial neural systems. For robotic applications, synaptic elements of such networks can rapidly acquire the kinematic invariances embedded within the presented samples. Subsequently, joint-space configurations, required to follow arbitrary end-effector trajectories, can readily be computed. In a significant departure from prior neuromorphic learning algorithms, this methodology provides mechanisms for incorporating an in-training skew to handle kinematics and environmental constraints.

Description: Computer literature searches were carried out at Duke University and NASA Langley Research Center. The purpose is to enhance personal knowledge based on the technical problems of pattern recognition and image understanding which must be solved for the Mars Rover and Sample Return Mission. Intensive study effort of a large collection of relevant literature resulted in a compilation of all important documents in one place. Furthermore, the documents are being classified into: Mars Rover; computer vision (theory); imaging systems; pattern recognition methodologies; and other smart techniques (AI, neural networks, fuzzy logic, etc).

Description: Neural networks have been applied to tasks in several areas of artificial intelligence, including vision, speech, and language. Relatively little work has been done in the area of problem solving. Two approaches to path-finding are presented, both using neural network techniques. Both techniques require a training period. Training under the back propagation (BPL) method was accomplished by presenting representations of (current position, goal position) pairs as input and appropriate actions as output. The Hebbian/interactive activation (HIA) method uses the Hebbian rule to associate points that are nearby. A path to a goal is found by activating a representation of the goal in the network and processing until the current position is activated above some threshold level. BPL, using back-propagation learning, failed to learn, except in a very trivial fashion, that is equivalent to table lookup techniques. HIA, performed much better, and required storage of fewer weights. In drawing a comparison, it is important to note that back propagation techniques depend critically upon the forms of representation used, and can be sensitive to parameters in the simulations; hence the BPL technique may yet yield strong results.

Description: An initial examination was conducted of an Intelligent Tutoring System (ITS) developed for use in industry. The ITS, developed by NASA, simulated a satellite deployment task. More specifically, the PD (Payload Assist Module Deployment)/ICAT (Intelligent Computer Aided Training) System simulated a nominal Payload Assist Module (PAM) deployment. The development of expertise on this task was examined using three Flight Dynamics Officer (FDO) candidates who had no previous experience with this task. The results indicated that performance improved rapidly until Trial 5, followed by more gradual improvements through Trial 12. The performance dimensions measured included performance speed, actions completed, errors, help required, and display fields checked. Suggestions for further refining the software and for deciding when to expose trainees to more difficult task scenarios are discussed. Further, the results provide an initial demonstration of the effectiveness of the PD/ICAT system in training the nominal PAM deployment task and indicate the potential benefits of using ITS's for training other FDO tasks.

Description: Analysis of a robot control system leads to a broad range of processing requirements. One fundamental requirement of a robot control system is the necessity of a microcomputer system in order to provide sufficient processing capability.The use of multiple processors in a parallel architecture is beneficial for a number of reasons, including better cost performance, modular growth, increased reliability through replication, and flexibility for testing alternate control strategies via different partitioning. A survey of the progression from low level control synchronizing primitives to higher level communication tools is presented. The system communication and control mechanisms of existing robot control systems are compared to the hierarchical control model. The impact of this design methodology on the current robot control systems is explored.

Description: Extended dialogues between a human user and a robot system are presented. The purpose of each dialogue is to teach the robot a new skill or to improve the performance of a skill it already has. The particular interest is in natural language dialogues but the illustrated techniques can be applied to any high level language. The primary purpose is to show how verbal instruction can be integrated with the robot's autonomous learning of a skill.

Description: The implementation of an earlier introduced neural net model for pattern classification is considered. Data flow principles are employed in the development of a machine that efficiently implements the model and can be useful for real time classification tasks. Further enhancement with optical computing structures is also considered.

Description: A system is proposed which receives input information from many sensors that may have diverse scaling, dimension, and data representations. The proposed system tolerates sensory information with faults. The proposed self-adaptive processing technique has great promise in integrating the techniques of artificial intelligence and neural networks in an attempt to build a more intelligent computing environment. The proposed architecture can provide a detailed decision tree based on the input information, information stored in a long-term memory, and the adapted rule-based knowledge. A mathematical model for analysis will be obtained to validate the cited hypotheses. An extensive software program will be developed to simulate a typical example of pattern recognition problem. It is shown that the proposed model displays attention, expectation, spatio-temporal, and predictory behavior which are specific to the human brain. The anticipated results of this research project are: (1) creation of a new dynamic neural network structure, and (2) applications to and comparison with conventional multi-layer neural network structures. The anticipated benefits from this research are vast. The model can be used in a neuro-computer architecture as a building block which can perform complicated, nonlinear, time-varying mapping from a multitude of input excitory classes to an output or decision environment. It can be used for coordinating different sensory inputs and past experience of a dynamic system and actuating signals. The commercial applications of this project can be the creation of a special-purpose neuro-computer hardware which can be used in spatio-temporal pattern recognitions in such areas as air defense systems, e.g., target tracking, and recognition. Potential robotics-related applications are trajectory planning, inverse dynamics computations, hierarchical control, task-oriented control, and collision avoidance.

Description: The problem of learned navigation of a circular robot R, of radius delta (is greater than or equal to 0), through a terrain whose model is not a-priori known is considered. Two-dimensional finite-sized terrains populated by an unknown (but, finite) number of simple polygonal obstacles are also considered. The number and locations of the vertices of each obstacle are unknown to R. R is equipped with a sensor system that detects all vertices and edges that are visible from its present location. In this context two problems are covered. In the visit problem, the robot is required to visit a sequence of destination points, and in the terrain model acquisition problem, the robot is required to acquire the complete model of the terrain. An algorithmic framework is presented for solving these two problems using a retraction of the freespace onto the Voronoi diagram of the terrain. Algorithms are then presented to solve the visit problem and the terrain model acquisition problem.

Description: Reactive schema-based navigation is possible in space domains by extending the methods developed for ground-based navigation found within the Autonomous Robot Architecture (AuRA). Reformulation of two dimensional motor schemas for three dimensional applications is a straightforward process. The manifold advantages of schema-based control persist, including modular development, amenability to distributed processing, and responsiveness to environmental sensing. Simulation results show the feasibility of this methodology for space docking operations in a cluttered work area.

Description: Adaptive systems under slow parameter adaption can be analyzed by the method of averaging. This provides a means to assess stability (and instability) properties of most adaptive systems, either continuous-time or (more importantly for practice) discrete-time, as well as providing an estimate of the region of attraction. Although the method of averaging is conceptually straightforward, even simple examples are well beyond hand calculations. Specific software tools are proposed which can provide the basis for user-friendly environment to perform the necessary computations involved in the averaging analysis.

Description: The general operation of KATE, an artificial intelligence controller, is outlined. A shuttle environmental control system (ECS) demonstration system for KATE is explained. The knowledge base model for this system is derived. An experimental test procedure is given to verify parameters in the model.

Description: The problem of automatic drawing was investigated in two ways. First, a DAB model of drawing processes was introduced. DAB stands for three types of knowledge hypothesized to support drawing abilities, namely, Drawing Knowledge, Assimilated Knowledge, and Base Knowledge. Speculation concerning the content and character of each of these subsystems of the drawing process is introduced and the overall adequacy of the model is evaluated. Second, eight experts were each asked to understand six engineering drawings and to think aloud while doing so. It is anticipated that a concurrent protocol analysis of these interviews can be carried out in the future. Meanwhile, a general description of the videotape database is provided. In conclusion, the DAB model was praised as a worthwhile first step toward solution of a difficult problem, but was considered by and large inadequate to the challenge of automatic drawing. Suggestions for improvements on the model were made.

Description: A major barrier in taking expert systems from prototype to operational status involves instilling end user confidence in the operational system. The software of different life cycle models is examined and the advantages and disadvantages of each when applied to expert system development are explored. The Fault Isolation Expert System for Tracking and data relay satellite system Applications (FIESTA) is presented as a case study of development of an expert system. The end user confidence necessary for operational use of this system is accentuated by the fact that it will handle real-time data in a secure environment, allowing little tolerance for errors. How FIESTA is dealing with transition problems as it moves from an off-line standalone prototype to an on-line real-time system is discussed.

Description: Artificial neural systems promise to integrate symbolic and sub-symbolic processing to achieve real time control of physical systems. Two potential alternatives exist. In one, neural nets can be used to front-end expert systems. The expert systems, in turn, are developed with varying degrees of parallelism, including their implementation in neural nets. In the other, rule-based reasoning and sensor data can be integrated within a single hybrid neural system. The hybrid system reacts as a unit to provide decisions (problem solutions) based on the simultaneous evaluation of data and rules. Discussed here is a model hybrid system based on the fuzzy cognitive map (FCM). The operation of the model is illustrated with the control of a hypothetical satellite that intelligently alters its attitude in space in response to an intersecting micrometeorite shower.

Description: A large scale domain-independent spatial data management expert system that serves as a front-end to databases containing spatial data is described. This system is unique for two reasons. First, it uses spatial search techniques to generate a list of all the primary keys that fall within a user's spatial constraints prior to invoking the database management system, thus substantially decreasing the amount of time required to answer a user's query. Second, a domain-independent query expert system uses a domain-specific rule base to preprocess the user's English query, effectively mapping a broad class of queries into a smaller subset that can be handled by a commercial natural language processing system. The methods used by the spatial search module and the query expert system are explained, and the system architecture for the spatial data management expert system is described. The system is applied to data from the International Ultraviolet Explorer (IUE) satellite, and results are given.

Description: Integrating diverse information sources and application software in a principled and general manner will require a very capable advanced information management (AIM) system. In particular, such a system will need a comprehensive addressing scheme to locate the material in its docuverse. It will also need a natural language processing (NLP) system of great sophistication. It seems that the NLP system must serve three functions. First, it provides an natural language interface (NLI) for the users. Second, it serves as the core component that understands and makes use of the real-world interpretations (RWIs) contained in the docuverse. Third, it enables the reasoning specialists (RSs) to arrive at conclusions that can be transformed into procedures that will satisfy the users' requests. The best candidate for an intelligent agent that can satisfactorily make use of RSs and transform documents (TDs) appears to be an object oriented data base (OODB). OODBs have, apparently, an inherent capacity to use the large numbers of RSs and TDs that will be required by an AIM system and an inherent capacity to use them in an effective way.

Description: Existing NASA supported scientific data bases are usually developed, managed and populated in a tedious, error prone and self-limiting way in terms of what can be described in a relational Data Base Management System (DBMS). The next generation Earth remote sensing platforms (i.e., Earth Observation System, (EOS), will be capable of generating data at a rate of over 300 Mbs per second from a suite of instruments designed for different applications. What is needed is an innovative approach that creates object-oriented databases that segment, characterize, catalog and are manageable in a domain-specific context and whose contents are available interactively and in near-real-time to the user community. Described here is work in progress that utilizes an artificial neural net approach to characterize satellite imagery of undefined objects into high-level data objects. The characterized data is then dynamically allocated to an object-oriented data base where it can be reviewed and assessed by a user. The definition, development, and evolution of the overall data system model are steps in the creation of an application-driven knowledge-based scientific information system.

Description: Applications of rapid prototyping and Artificial Intelligence techniques to problems associated with Space Station-era information management systems are described. In particular, the work is centered on issues related to: (1) intelligent man-machine interfaces applied to scientific data user support, and (2) the requirement that intelligent information management systems (IIMS) be able to efficiently process metadata updates concerning types of data handled. The advanced IIMS represents functional capabilities driven almost entirely by the needs of potential users. Space Station-era scientific data projected to be generated is likely to be significantly greater than data currently processed and analyzed. Information about scientific data must be presented clearly, concisely, and with support features to allow users at all levels of expertise efficient and cost-effective data access. Additionally, mechanisms for allowing more efficient IIMS metadata update processes must be addressed. The work reported covers the following IIMS design aspects: IIMS data and metadata modeling, including the automatic updating of IIMS-contained metadata, IIMS user-system interface considerations, including significant problems associated with remote access, user profiles, and on-line tutorial capabilities, and development of an IIMS query and browse facility, including the capability to deal with spatial information. A working prototype has been developed and is being enhanced.

Description: Earth Scientists lack adequate tools for quantifying complex relationships between existing data layers and studying and modeling the dynamic interactions of these data layers. There is a need for an earth systems tool to manipulate multi-layered, heterogeneous data sets that are spatially indexed, such as sensor imagery and maps, easily and intelligently in a single system. The system can access and manipulate data from multiple sensor sources, maps, and from a learned object hierarchy using an advanced knowledge-based geographical information system. A prototype Knowledge-Based Geographic Information System (KBGIS) was recently constructed. Many of the system internals are well developed, but the system lacks an adequate user interface. A methodology is described for developing an intelligent user interface and extending KBGIS to interconnect with existing NASA systems, such as imagery from the Land Analysis System (LAS), atmospheric data in Common Data Format (CDF), and visualization of complex data with the National Space Science Data Center Graphics System. This would allow NASA to quickly explore the utility of such a system, given the ability to transfer data in and out of KBGIS easily. The use and maintenance of the object hierarchies as polymorphic data types brings, to data management, a while new set of problems and issues, few of which have been explored above the prototype level.

Description: For practical automated diagnostic systems to continue functioning after failure, they must not only be able to diagnose sensor failures but also be able to tolerate the absence of data from the faulty sensors. It is shown that conventional (associational) diagnostic methods will have combinatoric problems when trying to isolate faulty sensors, even if they adequately diagnose other components. Moreover, attempts to extend the operation of diagnostic capability past sensor failure will necessarily compound those difficulties. Model-based reasoning offers a structured alternative that has no special problems diagnosing faulty sensors and can operate gracefully when sensor data is missing.

Description: The commonly stated truths that data may be segregated from program control in generic expert system shells and that such tools support straightforward knowledge representation were examined. The ideal of separation of data from program control in expert systems is difficult to realize for a variety of reasons. One approach to achieving this goal is to integrate hybrid collections of specialized shells and tools instead of producing custom systems built with a single all purpose expert system tool. Aspects of these issues are examined in the context of a specific diagnostic expert system application, the Backup Control Mode Analysis and Utility System (BCAUS), being developed for the Gamma Ray Observatory (GRO) spacecraft. The project and the knowledge gained in working on the project are described.

Description: Flexible robotic systems for space applications need to use local information to guide their action in uncertain environments where the state of the environment and even the goals may change. They have to be tolerant of unexpected events and robust enough to carry their task to completion. Tactical goals should be modified while maintaining strategic goals. Furthermore, reactive robotic systems need to have a broader view of their environments than sensory-based systems. An architecture and a theory of representation extending the basic cycles of action and perception are described. This scheme allows for dynamic description of the environment and determining purposive and timely action. Applications of this scheme for assembly and repair tasks using a Universal Machine Intelligence RTX robot are being explored, but the ideas are extendable to other domains. The nature of reactivity for sensor-based robotic systems and implementation issues encountered in developing a prototype are discussed.

Description: How a neural network can work, compared to a hybrid system based on an operations research and artificial intelligence approach, is investigated through a mission scheduling problem. The characteristic features of each system are discussed.

Description: The Ground Data Systems Resource Allocation Process at the Jet Propulsion Laboratory provides medium- and long-range planning for the use of Deep Space Network and Mission Control and Computing Center resources in support of NASA's deep space missions and Earth-based science. Resources consist of radio antenna complexes and associated data processing and control computer networks. A semi-automated system was developed that allows operations personnel to interactively generate, edit, and revise allocation plans spanning periods of up to ten years (as opposed to only two or three weeks under the manual system) based on the relative merit of mission events. It also enhances scientific data return. A software system known as the Resource Allocation and Planning Helper (RALPH) merges the conventional methods of operations research, rule-based knowledge engineering, and advanced data base structures. RALPH employs a generic, highly modular architecture capable of solving a wide variety of scheduling and resource sequencing problems. The rule-based RALPH system has saved significant labor in resource allocation. Its successful use affirms the importance of establishing and applying event priorities based on scientific merit, and the benefit of continuity in planning provided by knowledge-based engineering. The RALPH system exhibits a strong potential for minimizing development cycles of resource and payload planning systems throughout NASA and the private sector.

Description: The Multi-Satellite Operations Control Center (MSOCC) Jargon Interpreter (MJI) demonstrates an English language interface for a constrained domain. A constrained domain is defined as one with a small and well delineated set of actions and objects. The set of actions chosen for the MJI is from the domain of MSOCC Applications Executive (MAE) Systems Test and Operations Language (STOL) directives and contains directives for signing a cathode ray tube (CRT) on or off, calling up or clearing a display page, starting or stopping a procedure, and controlling history recording. The set of objects chosen consists of CRTs, display pages, STOL procedures, and history files. Translation from English sentences to STOL directives is done in two phases. In the first phase, an augmented transition net (ATN) parser and dictionary are used for determining grammatically correct parsings of input sentences. In the second phase, grammatically typed sentences are submitted to a forward-chaining rule-based system for interpretation and translation into equivalent MAE STOL directives. Tests of the MJI show that it is able to translate individual clearly stated sentences into the subset of directives selected for the prototype. This approach to an English language interface may be used for similarly constrained situations by modifying the MJI's dictionary and rules to reflect the change of domain.

Description: The advantages of an AI system of actively monitoring human control of a shared resource (such as a telerobotic manipulator) are presented. A system is described in which a simple AI planning program gains efficiency by monitoring human actions and recognizing when the actions cause a change in the system's assumed state of the world. This enables the planner to recognize when an interaction occurs between human actions and system goals, and allows maintenance of an up-to-date knowledge of the state of the world and thus informs the operator when human action would undo a goal achieved by the system, when an action would render a system goal unachievable, and efficiently replans the establishment of goals after human intervention.

Description: Space-based robots need low level control for collision detection and avoidance, short-term load management, fine-grained motion, and other physical tasks. In addition, higher level control is required to focus strategic decision making as missions are assigned and carried out. Reasoning and control must be responsive to ongoing changes in the environment. Research aimed at bridging the gap between high level artificial intelligence (AI) planning techniques and task-level robot programming for telerobotic systems is described. Situated reasoning is incorporated into AI and Robotics systems in order to coordinate a robot's activity within its environment. An integrated system under development in a component maintenance domain is described. It is geared towards replacing worn and/or failed Orbital Replacement Units (ORUs) designed for use aboard NASA's Space Station Freedom based on the collection of components available at a given time. High level control reasons in component space in order to maximize the number operational component-cells over time, while the task-level controls sensors and effectors, detects collisions, and carries out pick and place tasks in physical space. Situated reasoning is used throughout the system to cope with component failures, imperfect information, and unexpected events.

Description: NASA's Superfluid Helium On-Orbit Transfer (SHOOT) project is a Shuttle-based experiment designed to acquire data on the properties of superfluid helium in micro-gravity. Aft Flight Deck Computer Software for the SHOOT experiment is comprised of several monitoring programs which give the astronaut crew visibility into SHOOT systems and a rule based system which will provide process control, diagnosis and error recovery for a helium transfer without ground intervention. Given present Shuttle manifests, this software will become the first expert system to be used in space. The SHOOT Command and Monitoring System (CMS) software will provide a near real time highly interactive interface for the SHOOT principal investigator to control the experiment and to analyze and display its telemetry. The CMS software is targeted for all phases of the SHOOT project: hardware development, pre-flight pad servicing, in-flight operations, and post-flight data analysis.

Description: Developments and applications of artificial intelligence (AI) methods in the design of structural systems is reviewed. Principal shortcomings in the current approach are emphasized, and the need for some degree of formalism in the development environment for such design tools is underscored. Emphasis is placed on efforts to integrate algorithmic computations in expert systems.

Description: The technology for controlling elastic deformations of flexible structures is one of the key considerations for future space initiatives. A vital area needed to achieve this objective is the development of a control design methodology applicable to future structures. The mu-synthesis technique is employed to design a high-performance vibration attenuation controller for the JPL/AFAL experimental flexible antenna structure. The results presented deal primarily with the control of first two global flexible modes using only two hub actuators and two hub sensors. Implementation of the multivariable control laws based on a finite-element model is presented. All results are from actual implementation on the JPL/AFAL flexible structure testbed.

Description: An algorithm is developed for recursively calculating the minimum realization of a linear system from sampled impulse response data. The Gram-Schmidt orthonormalization technique is used to generate an orthonormal basis for factorization of the data matrix. The system matrix thus identified is in upper Hessenberg form, which has advantages for the identification of modal parameters including damping coefficients, frequencies, mode shapes, and modal participation factors. It also has the property that once an element of the system matrix is computed, it is never altered as the dimension of the model is increased in the recursive process. Numerical examples are presented for comparison of the recursive and nonrecursive forms of the eigensystem realization algorithm.

Description: A simple approach for controlling the manipulator configuration over the entire motion is presented, based on augmentation of the manipulator forward kinematics. User-defined kinematic functions and the end-effector Cartesian coordinates are combined to form a set of task-related configuration variables as generalized coordinates for the manipulator. A task-based adaptive scheme is then utilized to control the configuration variables and achieve tracking of the desired reference trajectories. This achieves the desired end-effector motion while utilizing redundancy to achieve any additional task. Simulation results for a direct-drive two-link arm are given to illustrate the proposed control scheme. The scheme has also been implemented for real-time control of three links of a PUMA 560 industrial robot. The simulation and experimental results validate the configuration control scheme and demonstrate its capabilities for performing various realistic tasks.

Description: Two issues that are fundamental to developing autonomous intelligent robots, namely, rudimentary learning capability and dexterous manipulation, are examined. A powerful neural learning formalism is introduced for addressing a large class of nonlinear mapping problems, including redundant manipulator inverse kinematics, commonly encountered during the design of real-time adaptive control mechanisms. Artificial neural networks with terminal attractor dynamics are used. The rapid network convergence resulting from the infinite local stability of these attractors allows the development of fast neural learning algorithms. Approaches to manipulator inverse kinematics are reviewed, the neurodynamics model is discussed, and the neural learning algorithm is presented.

Description: Comments are given on the application to rigid-link manipulators of geometric control theory, resolved acceleration control, operational space control, and nonlinear decoupling theory, and the essential unity of these techniques for externally linearizing and decoupling end-effector dynamics is discussed. Exploiting the fact that the mass matrix of a rigid-link manipulator is positive definite, and the fact that there is an independent input for each degree of freedom, it is shown that a necessary and sufficient condition for a locally externally linearizing and output decoupling feedback law to exist is that the end effector Jacobian matrix be nonsingular.

Description: An analytical model of the feedback control system which estimates the voltage generated by the piezoelectric sensor as a function of the dynamic stress at the sensor location and the force exerted by the driver piezoelectric as a function of signal gain is developed. The analytical results are compared to measured results for a cantilever beam excited to vibrate in its first natural mode. The estimated increase in the first mode damping factor is in good agreement with the measured results.

Description: A method is investigated for optimizing the end-to-end performance of image gathering and restoration for visual quality. To achieve this objective, one must inevitably confront the problems that the visual quality of restored images depends on perceptual rather than mathematical considerations and that these considerations vary with the target, the application, and the observer. The method adopted in this paper is to optimize image gathering informationally and to restore images interactively to obtain the visually preferred trade-off among fidelity resolution, sharpness, and clarity. The results demonstrate that this method leads to significant improvements in the visual quality obtained by the traditional digital processing methods. These traditional methods allow a significant loss of visual quality to occur because they treat the design of the image-gathering system and the formulation of the image-restoration algorithm as two separate tasks and fail to account for the transformations between the continuous and the discrete representations in image gathering and reconstruction.

Description: The control-structure interaction problem is addressed via stability analysis of a generic linear servo loop model. With the plant described by the rigid body mode and a single elastic mode, structural flexibility is categorized into one of three types: (1) appendage, (2) in-the-loop minimum phase, and (3) in-the-loop nonminimum phase. Closing the loop with proportional-derivative (PD) control action and introducing sensor roll-off dynamics in the feedback path, stability conditions are obtained. Trade studies are conducted with modal frequency, modal participation, modal damping, loop bandwidth, and sensor bandwidth treated as free parameters. Results indicate that appendage modes are most likely to produce instability if they are near the sensor rolloff, whereas in-the-loop modes are most dangerous near the loop bandwidth. The main goal of this paper is to provide a fundamental understanding of the control-structure interaction problem so that it may benefit the design of complex spacecraft and pointing system servo loops. In this framework, the JPL Pathfinder gimbal pointer is considered as an example.

Description: A cortical neural network that computes the visibility of shifts in the direction of movement is proposed. The network computes: (1) the magnitude of the position difference between the test and background patterns, (2) localized contrast differences at different spatial scales analyzed by computing temporal gradients of the difference and sum of the outputs of paired even- and odd-symmetric bandpass filters convolved with the input pattern, and (3) using global processes that pool the output from paired even- and odd-symmetric simple and complex cells across the spatial extent of the background frame of reference the direction a test pattern moved relative to a textured background. Evidence that magnocellular pathways are used to discriminate the direction of movement is presented. Since magnocellular pathways are used to discriminate the direction of movement, this task is not affected by small pattern changes such as jitter, short presentations, blurring, and different background contrasts that result when the veiling illumination in a scene changes.

Description: The present treatment of the related problems of minimal inversion and perfect output control in linear multivariable systems uses a simple analytical expression for the inverse of a square multivariate system's transfer-function matrix to construct a minimal-order inverse of the system. Because the poles of the minimal-order inverse are the transmission zeros of the system, necessary and sufficient conditions for the inverse system's stability are simply stated in terms of the zero polynomial of the original system. A necessary and sufficient condition for the existence of the required controllers is that the plant zero polynomial be neither identical to zero nor unstable.

Description: A state reconstructor for deterministic digital systems is presented which is ideal in the following sense: if the plant parameters are known exactly, the output of the state reconstructor will exactly equal the true state of the plant, not just approximate it. Furthermore, this ideal state reconstructor adds no additional states or eigenvalues to the system. Nor does it affect the plant equation for the system in any way; it affects only the measurement equation. While there are countless ways of choosing the ideal state reconstructor parameters, two distinct methods are described here. An example is presented which illustrates the procedures to completely design the ideal state reconstructor using both methods.

Description: It is proposed that chaotic attractors incorporated in neural net models can represent classes of patterns in the same way in which a set of static attractors represent unrelated patterns. Therefore, chaotic states of neuron activity are associated with higher level cognitive processes such as generalization and abstraction.

Description: The Dryden Flight Research Facility of the NASA Ames Research Facility of the NASA Ames Research Center is developing a rapid prototyping facility for flight research in flight systems concepts that are based on artificial intelligence (AI). The facility will include real-time high-fidelity aircraft simulators, conventional and symbolic processors, and a high-performance research aircraft specially modified to accept commands from the ground-based AI computers. This facility is being developed as part of the NASA-DARPA automated wingman program. This document discusses the need for flight research and for a national flight research facility for the rapid prototyping of AI-based avionics systems and the NASA response to those needs.

Description: The convergence mechanism of vectors in Hopfield's neural network in relation to recognition of partially known patterns is studied in terms of both inner products and Hamming distance. It has been shown that Hamming distance should not always be used in determining the convergence of vectors. Instead, inner product weighting coefficients play a more dominant role in certain data representations for determining the convergence mechanism. A trinary neuron representation for associative memory is found to be more effective for associative recall. Applications of the trinary associative memory to reconstruct machine part images that are partially missing are demonstrated by means of computer simulation as examples of the usefulness of this approach.

Description: NASA plans to solve some of the problems of handling large-scale scientific data bases by turning to artificial intelligence (AI) are discussed. The growth of the information glut and the ways that AI can help alleviate the resulting problems are reviewed. The employment of the Intelligent User Interface prototype, where the user will generate his own natural language query with the assistance of the system, is examined. Spatial data management, scientific data visualization, and data fusion are discussed.

Description: The transmission zeros of a large space structure are frequently computed by means of the general-purpose algorithm of Emami-Naeini and Van Dooren (1982). It is shown that careful exploitation of the special form of the equations of motion of structural dynamics leads to an algorithm that is at least 60 times as fast as this when applied to an undamped structure, and 15 times as fast for a lightly damped one.

Description: A simple method is presented for combined signal synthesis and parameter adaptation within the framework of model reference adaptive control theory. The results are obtained using a simple derivation based on an improved Liapunov function.

Description: The author presents a simple decentralized adaptive-control scheme for multijoint robot manipulators based on the independent joint control concept. The control objective is to achieve accurate tracking of desired joint trajectories. The proposed control scheme does not use the complex manipulator dynamic model, and each joint is controlled simply by a PID (proportional-integral-derivative) feedback controller and a position-velocity-acceleration feedforward controller, both with adjustable gains. Simulation results are given for a two-link direct-drive manipulator under adaptive independent joint control. The results illustrate trajectory tracking under coupled dynamics and varying payload. The proposed scheme is implemented on a MicroVAX II computer for motion control of the three major joints of a PUMA 560 arm. Experimental results are presented to demonstrate that trajectory tracking is achieved despite coupled nonlinear joint dynamics.

Description: The article presents a new and simple solution to the obstacle avoidance problem for redundant robots. In the proposed approach, called configuraiton control, the redundancy is utilized to configure the robot so as to satisfy a set of kinematic inequality constraints representing obstacle avoidance, while the end-effector is tracking a desired trajectory. The robot control scheme is very simple, and uses on-line adaptation to eliminate the need for the complex dynamic model and parameter values of the robot. Several simulation results for a four-link planar robot are presented to illustrate the versatility of the approach. These include reaching around a stationary obstacle, simultaneous avoidance of two obstacles, robot reconfiguration to avoid a moving obstacle, and avoidance of rectangular obstacles.

Description: Theoretical and implementation aspects of AI systems for space applications are discussed in reviews and reports. Sections are devoted to planning and scheduling, fault isolation and diagnosis, data management, modeling and simulation, and development tools and methods. Particular attention is given to a situated reasoning architecture for space repair and replace tasks, parallel plan execution with self-processing networks, the electrical diagnostics expert system for Spacelab life-sciences experiments, diagnostic tolerance for missing sensor data, the integration of perception and reasoning in fast neural modules, a connectionist model for dynamic control, and applications of fuzzy sets to the development of rule-based expert systems.

Description: Intelligence robotic system design involves sensor allocation and sensor data fusion. In the field of geometric reasoning, a fast and effective representation method is needed to determine the physical dimensions of the manipulator as well as the motion environment in sufficient detail. Due to the capability of current sensor technology, the integrity of the sensor information is location dependent and point-wise, line-wise and plane-wise in nature. In order to maneuver the robot by reasoning the environment, there is a minimum requirement of prior information for integrating the allocated sensor feedback. Some new theoretical developments of how to associate the number and the locations of the sensors with three-dimensional geometry of the robot and the motion environment is presented. The related collision avoidance problem (CAP) will be addressed as well.

Description: It is shown that vertex implication results in parameter space apply to interval trigonometric polynomials, which characterize the frequency response of a finite impulse response (FIR) digital filter. A weak result and a strong result are given, and the conclusions are illustrated by examples.

Description: An estimate is made of the motion parameters, namely, linear and angular velocities, of a rigid body rotating and translating in three-space. The authors assume that the velocities are constant and that the motion is not completely observable. They consider two separate cases of partial observations corresponding to the orthogonal and the perspective projections, respectively. If (x, y, z) is the Cartesian coordinate of the three-space, the authors assume in the first case that the projection of the motion on the x-y plane is observed. If (r, theta, phi) is the polar coordinates of the three-space, they assume in the second case that the parameter vector (theta, phi) is observed. The use of both of these cases to estimate the motion parameters is discussed.