Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Learning”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 523 records · Page 29

PALMO: An OVERFLOW Machine Learning Airfoil Performance Database

The OVERFLOW Machine Learning Airfoil Performance (PALMO) database has been created to enable robust modeling of airfoil performance in a variety of applications. The PALMO database uses OVERFLOW simulation data second-order accurate in time and fourth-order accurate in space with Spalart-Allmaras turbulence closure. The foundation of the in-development PALMO database is the airfoil base cube. Each base cube includes simulation data parametrized over a range of Mach numbers, Reynolds numbers, and angles-of-attack. This database includes the NACA 4-series airfoils, with parametrization in airfoil thickness and camber from an NACA 0006 to an NACA 4424. In total, 52,480 NACA 4-series OVERFLOW calculations were run on the NASA High-End Compute Capability (HECC) supercomputer. This provides high-order-accurate simulation data covering a wide range of aerospace design applications, which enables users to develop accurate airfoil performance look-up tables without additional high-performance computing. In addition to engineering design and analysis of aerospace vehicles, PALMO is well suited to be a benchmark dataset for the development and testing of machine learning methods in aerospace engineering. This work presents an example PALMO surrogate model that enables accurate airfoil performance predictions for any arbitrary combination of camber, thickness, Mach number, Reynolds number, and angle of attack within the bounds of the database. Airfoil performance tables predicted for an airfoil not used in training the model are used in three-dimensional OVERFLOW simulations to quantify the downstream accuracy on aggregate rotor performance metrics. For the NACA 3415 airfoil, which had no common thickness or camber with the training data, the surrogate predicted and CFD generated tables were within 2.1% of each other in the forward flight lift to drag metric. This suggests that performance tables generated for airfoils within the bounds of the PALMO database will yield aggregate rotor performance predictions on par with tables generated from directly running OVERFLOW airfoil calculations. The PALMO airfoil performance coefficients are available publicly.

Database↗

Machine Learning Application for Improving Cloud Detection and Phase Determination Over Sunglint Regions for Geostationary Satellites

Cloud detection and phase determination over sunglint regions has been a challenge, especially for geostationary (GEO) satellites. Sunglint is observed when the sunlight specular reflection is at the same viewing angle of the satellite sensor. This intense reflection in the visible channels (VIS) is often comparable to that from optically thick clouds. It also contaminates the shortwave infrared channels (SWIR). Consequently, VIS and SWIR channels become less useful - or not useful- when they are saturated, hampering the detection of cloudy and clear-sky pixels. Sunglint contamination happens frequently and exists nearly in every daytime GEO full disk satellite images. However, sunglint intensity and region are difficult to model due to variable viewing geometry and ocean surface conditions. Moreover, existing physical models do not meet the accuracy required for operational GEO satellite cloud detection. We developed a machine learning algorithm to improve cloud detection in sunglint conditions for the NASA Langley’s Satellite ClOud and radiation Property retrieval System (SatCORPS). This poster presents our recent progress in the algorithm development, validation and applications. The algorithm is validated using collocated SatCORPS GOES-East and GOES-West cloud products. We demonstrate that the machine learning cloud detection in sunglint regions is superior to the traditional approach by improving temporal consistency between sunglint and non-sunglint conditions.

Machine Learning, Cloud detection, Sunglint, SatCO↗

Learning with a probabilistic teacher

Learning scheme for solving unsupervised learning problems with correct estimate convergence and for state estimates of Gauss-Markov sequences with additive and multiplicative observed noise

Agrawala, A. K.↗

Congruent and spurious motion in the learning and performance of a compensatory tracking task.

Examination of the importance of congruent and spurious yaw motion in compensatory tracking by eight airline pilots. The pilots, seated erect in the Ames man-carrying rotation device(MCRD), tracked with k/s+1 and k/s(s+1) vehicle dynamics in fixed- and moving-base simulation. Following the learning phase of the experiment, five levels of spurious angular acceleration were superimposed on the motion of the MCRD. Learning of the tracking task was found to be a function of both vehicle dynamics and mode of simulation. The presence of congruent motion information reduced learningtime in k/s(s+1) vehicle dynamics and resulted in lower tracking error in both vehicle dynamics. The spurious angular accelerations resulted in an increase in pilot tracking error; however, the relationship between the magnitude of the acceleration and its effect was highly complex. The data suggest that the minimal disturbance level for spurious angular accelerations during tracking is below 0.4 degec.

Guercio, J. G.↗

Robot learning and error correction

A model of robot learning is described that associates previously unknown perceptions with the sensed known consequences of robot actions. For these actions, both the categories of outcomes and the corresponding sensory patterns are incorporated in a knowledge base by the system designer. Thus the robot is able to predict the outcome of an action and compare the expectation with the experience. New knowledge about what to expect in the world may then be incorporated by the robot in a pre-existing structure whether it detects accordance or discrepancy between a predicted consequence and experience. Errors committed during plan execution are detected by the same type of comparison process and learning may be applied to avoiding the errors.

Friedman, L.↗

Vehicle Steering control: A model of learning

A hierarchy of strategies were postulated to describe the process of learning steering control. Vehicle motion and steering control data were recorded for twelve novices who drove an instrumented car twice a week during and after a driver training course. Car-driver describing functions were calculated, the probable control structure determined, and the driver-alone transfer function modelled. The data suggested that the largest changes in steering control with learning were in the way the driver used the lateral position cue.

Smiley, A.↗

Learning with imperfectly labeled patterns

The problem of learning in pattern recognition using imperfectly labeled patterns is considered. The performance of the Bayes and nearest neighbor classifiers with imperfect labels is discussed using a probabilistic model for the mislabeling of the training patterns. Schemes for training the classifier using both parametric and non parametric techniques are presented. Methods for the correction of imperfect labels were developed. To gain an understanding of the learning process, expressions are derived for success probability as a function of training time for a one dimensional increment error correction classifier with imperfect labels. Feature selection with imperfectly labeled patterns is described.

Chittineni, C. B.↗

Lessons learned - Americans in space

Milestones in the learning process during development of spaceflight capability for different NASA missions are traced from the initial launch of the Mercury escape rocket to the advent of the Shuttle. A detached heat shield on the first manned Mercury orbital flight led to a decision for reentry with retrorockets attached, a situation which demonstrated a need for on-line communications with the spacecraft construction engineers. The Gemini program served to develop EVA techniques and docking procedures essential for the Apollo missions. Apollo 8 circled the moon after only one manned test flight of the equipment due to the success of the first flight. Lunar landing simulation training exposed the astronauts to situations beyond the human ability to respond. The training was justified when computer overload required the pilot to retake manual control of some spacecraft functions during the successful first lunar landing. All lessons previously learned were used to safely return the damaged Apollo 13 to earth. The Skylab missions provided an opportunity to maintain rotating teams of flight controllers on a 24-hr/day basis for periods of months. Finally, the six-year hiatus of U.S. manned space activities between the end of the Skylab missions and the first Shuttle flight provided time for incorporating previous lessons into new procedures, training strategies, and data resources.

Kranz, E.↗

The Impact of Pictorial Display on Operator Learning and Performance

The effects of pictorially displayed information on human learning and performance of a simple control task were investigated. The controlled system was a harmonic oscillator and the system response was displayed to subjects as either an animated pendulum or a horizontally moving dot. Results indicated that the pendulum display did not effect performance scores but did significantly effect the learning processes of individual operators. The subjects with the pendulum display demonstrated more vertical internal models early in the experiment and the manner in which their internal models were tuned with practice showed increased variability between subjects.

Miller, R. A.↗

Learned pattern recognition using synthetic-discriminant-functions

A method of using synthetic-discriminant-functions to facilitate learning in a pattern recognition system is discussed. Learning is accomplished by continually adding images to the training set used for synthetic discriminant functions (SDF) construction. Object identification is performed by efficiently searching a library of SDF filters for the maximum optical correlation. Two library structures are discussed - binary tree and multilinked graph - along with maximum ascent, back-tracking, perturbation, and simulated annealing searching techniques. By incorporating the distortion invariant properties of SDFs within a library structure, a robust pattern recognition system can be produced.

Jared, David A.↗

Learning control for robotic manipulators with sparse data

Learning control algorithms have been proposed for error compensation in repetitive robotic manipulator tasks. It is shown that the performance of such control algorithms can be seriously degraded when the feedback data they use is relatively sparse in time, such as might be provided by vision systems. It is also shown that learning control algorithms can be modified to compensate for the effects of sparse data and thereby yield performance which approaches that of systems without limitations on the sensory information available for control.

Morita, Atsushi↗

Performance improvement of robots using a learning control scheme

Many applications of robots require that the same task be repeated a number of times. In such applications, the errors associated with one cycle are also repeated every cycle of the operation. An off-line learning control scheme is used here to modify the command function which would result in smaller errors in the next operation. The learning scheme is based on a knowledge of the errors and error rates associated with each cycle. Necessary conditions for the iterative scheme to converge to zero errors are derived analytically considering a second order servosystem model. Computer simulations show that the errors are reduced at a faster rate if the error rate is included in the iteration scheme. The results also indicate that the scheme may increase the magnitude of errors if the rate information is not included in the iteration scheme. Modification of the command input using a phase and gain adjustment is also proposed to reduce the errors with one attempt. The scheme is then applied to a computer model of a robot system similar to PUMA 560. Improved performance of the robot is shown by considering various cases of trajectory tracing. The scheme can be successfully used to improve the performance of actual robots within the limitations of the repeatability and noise characteristics of the robot.

Krishna, Ramuhalli↗

Production system chunking in SOAR: Case studies in automated learning

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.

Allen, Robert↗

Neuromorphic learning of continuous-valued mappings from noise-corrupted data. Application to real-time adaptive control

The ability of feed-forward neural network architectures to learn continuous valued mappings in the presence of noise was demonstrated in relation to parameter identification and real-time adaptive control applications. An error function was introduced to help optimize parameter values such as number of training iterations, observation time, sampling rate, and scaling of the control signal. The learning performance depended essentially on the degree of embodiment of the control law in the training data set and on the degree of uniformity of the probability distribution function of the data that are presented to the net during sequence. When a control law was corrupted by noise, the fluctuations of the training data biased the probability distribution function of the training data sequence. Only if the noise contamination is minimized and the degree of embodiment of the control law is maximized, can a neural net develop a good representation of the mapping and be used as a neurocontroller. A multilayer net was trained with back-error-propagation to control a cart-pole system for linear and nonlinear control laws in the presence of data processing noise and measurement noise. The neurocontroller exhibited noise-filtering properties and was found to operate more smoothly than the teacher in the presence of measurement noise.

Troudet, Terry↗

Computational neural learning formalisms for manipulator inverse kinematics

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.

Gulati, Sandeep↗

Learning in stochastic neural networks for constraint satisfaction problems

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.

Johnston, Mark D.↗

Integrating planning, execution, and learning

To achieve the goal of building an autonomous agent, the usually disjoint capabilities of planning, execution, and learning must be used together. An architecture, called MAX, within which cognitive capabilities can be purposefully and intelligently integrated is described. The architecture supports the codification of capabilities as explicit knowledge that can be reasoned about. In addition, specific problem solving, learning, and integration knowledge is developed.

Kuokka, Daniel R.↗

Reactive behavior, learning, and anticipation

Reactive systems always act, thinking only long enough to 'look up' the action to execute. Traditional planning systems think a lot, and act only after generating fairly precise plans. Each represents an endpoint on a spectrum. It is argued that primitive forms of reasoning, like anticipation, play an important role in reducing the cost of learning and that the decision to act or think should be based on the uncertainty associated with the utility of executing an action in a particular situation. An architecture for an adaptable reactive system is presented and it is shown how it can be augmented with a simple anticipation mechanism that can substantially reduce the cost and time of learning.

Whitehead, Steven D.↗