Search NASA⌕ Search

SEARCH · Search NASA

Results for “Model predictive control”

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 1,459 records · Page 81

Anomalous Hall crystals in rhombohedral multilayer graphene. II. General mechanism and a minimal model

Here we propose a minimal "three-patch model"for the anomalous Hall crystal (AHC), a topological electronic state that spontaneously breaks both time-reversal symmetry and continuous translation symmetry. The proposal for this state is inspired by the recently observed integer and fractional quantum Hall states in rhombohedral multilayer graphene at zero magnetic field. There, interaction effects appear to amplify the effects of a weak moiré potential, leading to the formation of stable, isolated Chern bands. It has been further shown that Chern bands are stabilized in mean-field calculations even without a moiré potential, enabling a realization of the AHC state. Our model is built on the dissection of the Brillouin zone into patches centered around high-symmetry points. Within this model, the wave functions at high-symmetry points fully determine the topology and energetics of the state. We extract two quantum geometrical phases of the noninteracting wave functions that control the stability of the topologically nontrivial AHC state. The model predicts that the AHC state wins over the topological trivial Wigner crystal in a wide range of parameters, and agrees very well with the results of full self-consistent Hartree-Fock calculations of the rhombohedral multilayer graphene Hamiltonian.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Expected individual benefit of prophylactic platelet transfusions in hemato‐oncology patients based on bleeding risks

Abstract Background Prophylactic platelet transfusions prevent bleeding in hemato‐oncology patients, but it is unclear how any benefit varies between patients. Our aim was to assess if patients with different baseline risks for bleeding benefit differently from a prophylactic platelet transfusion strategy. Study design and methods Using the data from the randomized controlled TOPPS trial (Trial of Platelet Prophylaxis), we developed a prediction model for World Health Organization grades 2, 3, and 4 bleeding risk (defined as at least one bleeding episode in a 30 days period) and grouped patients in four risk‐quartiles based on this predicted baseline risk. Predictors in the model were baseline platelet count, age, diagnosis, disease modifying treatment, disease status, previous stem cell transplantation, and the randomization arm. Results The model had a c‐statistic of 0.58 (95% confidence interval [CI] 0.54–0.64). There was little variation in predicted risks (quartiles 46%, 47%, and 51%), but prophylactic platelet transfusions gave a risk reduction in all risk quartiles. The absolute risk difference (ARD) was 3.4% (CI −12.2 to 18.9) in the lowest risk quartile (quartile 1), 7.4% (95% CI −8.4 to 23.3) in quartile 2, 6.8% (95% CI −9.1 to 22.9) in quartile 3, and 12.8% (CI −3.1 to 28.7) in the highest risk quartile (quartile 4). Conclusion In our study, generally accepted bleeding risk predictors had limited predictive power (expressed by the low c‐statistic), and, given the wide confidence intervals of predicted ARD, could not aid in identifying subgroups of patients who might benefit more (or less) from prophylactic platelet transfusion.

Cornelissen, Loes L.↗

Catalyst Ammonia Storage Measurements Using Radio Frequency Sensing1

Motivated by increasingly strict nitrogen oxides (NOx) limits, engine manufacturers have adopted selective catalytic reduction (SCR) technology to reduce engine-out NOx. In the SCR process, NOx react with ammonia (NH3) to form nitrogen and water vapor. The reaction is influenced by several variables, including stored ammonia on the catalyst, exhaust gas composition, and catalyst temperature. Currently, measurements from NOx and/or NH3 sensors upstream and downstream of the SCR are used with predictive models to estimate ammonia storage levels on the catalyst and control urea dosing. This study investigated a radio frequency (RF)-based method to directly monitor the ammonia storage state of the SCR. This approach utilizes the catalyst as a cavity resonator, in which an RF antenna excites electromagnetic waves within the cavity to monitor changes in the catalyst state. Ammonia storage causes changes in the dielectric properties of the catalyst, which directly impacts the RF signal. Changes in the RF signal relative to stored ammonia (NH3) were evaluated over a wide range of frequencies, temperatures, and exhaust conditions. The RF response to NH3 storage, desorption, and oxidation on the SCR was well correlated with changes in the catalyst state. Calibrated RF measurements demonstrate the ability to monitor the adsorption state of the SCR to within 10% of the sensor full scale. The results indicate direct measurement of SCR ammonia storage levels, and resulting catalyst feedback control, via RF sensing to have significant potential for optimizing the SCR system to improve NOx conversion and decrease urea consumption.

Engineering↗

CATALYST AMMONIA STORAGE MEASUREMENTS USING RADIO FREQUENCY SENSING

Motivated by increasingly strict NO x limits, engine manufactures have adopted selective catalytic reduction (SCR) technology to reduce engine-out NO x below mandated levels. In the SCR process, nitrogen oxides (NO x ) react with ammonia (NH 3 ) to form nitrogen and water vapor. The reaction is influenced by several variables, including stored ammonia on the catalyst, exhaust gas composition, and catalyst temperature. Currently, measurements from NO x and/or NH 3 sensors upstream and downstream of the SCR are used with predictive models to estimate ammonia storage levels on the catalyst and control urea dosing. This study investigated a radio frequency (RF) -based method to directly monitor the ammonia storage state of the SCR catalyst. This approach utilizes the SCR catalyst as a cavity resonator, in which an RF antenna excites electromagnetic waves within the cavity to monitor changes in the catalyst state. Ammonia storage causes changes in the dielectric properties of the catalyst, which directly impacts the RF signal. Changes in the RF signal relative to stored ammonia (NH 3 ) were evaluated over a wide frequency range as well as temperature and exhaust conditions. The RF response to NH 3 storage, desorption, and oxidation on the SCR was observed to be well-correlated with changes in the catalyst state. Calibrated RF measurements demonstrate the ability to monitor the adsorption state of the SCR to within 10 % of the sensor full scale. The results indicate direct measurement of SCR ammonia storage levels, and resulting catalyst feedback control, via RF sensing to have significant potential for optimizing the SCR system to improve NO x conversion and decrease urea consumption.

Aguilar, Jonathan↗

Model Development for Current–Voltage and Transconductance Characteristics of Normally-off AlN/GaN MOSHEMT

In this paper, an AlN/GaN-based MOSHEMT is proposed, in accordance to this, a charge control model has been developed analytically and simulated with MATLAB to predict the characteristics of threshold voltage, drain currents and transconductance. The physics based models for 2DEG density, threshold voltage and quantum capacitance in the channel has been put forward. By using these developed models, the drain current for both linear and saturation models is derived. The predicted threshold voltage with the variation of barrier thickness has been plotted. A positive threshold voltage can be obtained by decreasing the barrier thickness which builds up the foundation for enhancement mode MOSHEMT devices. The predicted I{sub d}–V{sub gs}, I{sub d}–V{sub ds} and transconductance characteristics show an excellent agreement with the experimental results and hence validate the model.

36 MATERIALS SCIENCE↗

Transcript Analysis for Internal Biodosimetry Using Peripheral Blood from Neuroblastoma Patients Treated with 131 I-mIBG, a Targeted Radionuclide

Calculating internal dose from therapeutic radionuclides currently relies on estimates made from multiple radiation exposure measurements, converted to absorbed dose in specific organs using the Medical Internal Radiation Dose (MIRD) schema. As an alternative biodosimetric approach, we utilized gene expression analysis of whole blood from patients receiving targeted radiotherapy. Collected blood from patients with relapsed or refractory neuroblastoma who received 131 I-labeled metaiodobenzylguanidine ( 131 I-mIBG) at the University of California San Francisco (UCSF) was used to compare calculated internal dose with the modulation of chosen gene expression. A total of 40 patients, median age 9 years, had blood drawn at baseline, 72 and 96 h after 131 I-mIBG infusion. Whole-body absorbed dose was calculated for each patient based on the cumulated activity determined from injected mIBG activity and patient-specific time-activity curves combined with 131 I whole-body S factors. We then assessed transcripts that were the most significant for describing the mixed therapeutic treatments over time using real-time polymerase chain reaction (RT-PCR). Modulation was evaluated statistically using multiple regression analysis for data at 0, 72 and 96 h. A total of 10 genes were analyzed across 40 patients: CDKN1A; FDXR; GADD45A; BCLXL; STAT5B; BAX; BCL2; DDB2; XPC; and MDM2. Six genes were significantly modulated upon exposure to 131 I-mIBG at 72 h, as well as at 96 h. Four genes varied significantly with absorbed dose when controlling for time. A gene expression biodosimetry model was developed to predict absorbed dose based on modulation of gene transcripts within whole blood. Three transcripts explained over 98% of the variance in the modulation of gene expression over the 96 h (CDKN1A, BAX and DDB2). To our knowledge, this is a novel study, which uses whole blood collected from patients treated with a radiopharmaceutical, to characterize biomarkers that may be useful for biodosimetry. Our data indicate that transcripts, which have been previously identified as biomarkers of external exposures in ex vivo whole blood and in vivo radiotherapy patients, are also good early indicators of internal exposure. However, for internal sources of radiation, the biokinetics and physical decay of the radionuclide strongly influence the gene expression.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Griffin Capability Improvements in Support of Ex-core Deep-Penetration Problems

Advanced reactor designs, especially portable reactors that are designed to be located closer to humans and operate autonomously, require the ability to accurately compute the ex-core neutron and gamma flux solutions in terms of shielding design optimization to reduce dose rates at the vessel boundary and detector signal prediction to drive the reactor control system. The Nuclear Energy Advanced Modeling and Simulation program has prioritized improvements to the Griffin discrete ordinates (SN) solver for deep-penetration problems in fiscal year 2025. Significant advancements have been made to the Griffin methodologies for solving ex-core deep-penetration problems for steady-state, fixed-source and transient calculations. This work presents the methodology improvements as well as a comprehensive demonstration with a Transient Test Reactor model and measurements.

22 - GENERAL STUDIES OF NUCLEAR REACTORS↗

SLUDGE WASHING AND DEMONSTRATION OF THE DWPF FLOWSHEET IN THE SRNL SHIELDED CELLS FOR SLUDGE BATCH 6 QUALIFICATION

Prior to initiating a new sludge batch in the Defense Waste Processing Facility (DWPF), Savannah River National Laboratory (SRNL) is required to simulate this processing, including Chemical Process Cell (CPC) simulation, waste glass fabrication, and chemical durability testing. This report documents this simulation for the next sludge batch, Sludge Batch 6 (SB6). SB6 consists of Tank 12 material that has been transferred to Tank 51 and subjected to Low Temperature Aluminum Dissolution (LTAD), Tank 4 sludge, and H-Canyon Pu solutions. Following LTAD and the Tank 4 addition, Liquid Waste Operations (LWO) provided SRNL a 3 L sample of Tank 51 sludge for SB6 qualification. Pu solution from H Canyon was also received. SB6 qualification included washing the sample per LWO plans/projections (including the addition of Pu from H Canyon), DWPF CPC simulations, waste glass fabrication (vitrification), and waste glass characterization and chemical durability evaluation. The following are significant observations from this demonstration. Sludge settling improved slightly as the sludge was washed. SRNL recommended (and the Tank Farm implemented) one less wash based on evaluations of Tank 40 heel projections and projections of the glass composition following transfer of Tank 51 to Tank 40. Thorium was detected in significant quantities (>0.1 wt % of total solids) in the sludge. In past sludge batches, thorium has been determined by Inductively Coupled Plasma-Mass Spectroscopy (ICP-MS), seen in small quantities, and reported with the radionuclides. As a result of the high thorium, SRNL-AD has added thorium to their suite of Inductively Coupled Plasma-Atomic Emission Spectroscopy (ICP-AES) elements. The acid stoichiometry for the DWPF Sludge Receipt and Adjustment Tank (SRAT) processing of 115%, or 1.3 mol acid per liter of SRAT receipt slurry, was adequate to accomplish some of the goals of SRAT processing: nitrite was destroyed to below 1,000 mg/kg and mercury was removed to below the DWPF target with 750 g of steam per g of mercury. However, rheological properties did not improve and were above the design basis. Hydrogen generation rates did not exceed DWPF limits during the SRAT and Slurry Mix Evaporator (SME) cycles. However, hydrogen generation during the SRAT cycle approached the DWPF limit. The glass fabricated with the Tank 51 SB6 SME product and Frit 418 was acceptable with respect to chemical durability as measured by the Product Consistency Test (PCT). The PCT response was also predictable by the current durability models of the DWPF Product Composition Control System (PCCS). It should be noted, however, that in the first attempt to make glass from the SME product, the contents of the fabrication crucible foamed over. This may be a result of the SME product's REDOX (Reduction/Oxidation - Fe{sup 2+}/{Sigma}Fe) of 0.08 (calculated from SME product analytical results). The following are recommendations drawn from this demonstration. In this demonstration, at the request of DWPF, SRNL caustic boiled the SRAT contents prior to acid addition to remove water (to increase solids concentration). During the nearly five hours of caustic boiling, 700 ppm of antifoam was required to control foaming. SRNL recommends that DWPF not caustic boil/concentrate SRAT receipt prior to acid addition until further studies can be performed to provide a better foaming control strategy or a new antifoam is developed for caustic boiling. Based on this set of runs and a recently completed demonstration with the SB6 Waste Acceptance Product Specifications (WAPS) sample, it is recommended that DWPF not add formic acid at the design addition rate of two gallons per minute for this sludge batch. A longer acid addition time appears to be helpful in allowing slower reaction of formic acid with the sludge and possibly decreases the chance of a foam over during acid addition.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Advancing Replicable Solutions for High-Performance Homes in the Southeast - JMC - Patrick Square

The work presented in this report advances the goals of the U.S. Department of Energy Building America program by improving the energy peformance of affordable and market-rate housing. Southface Energy Institute (Southface), part of the U.S. Department of Energy Building America research team Partnership for Home Innovation, worked with owners and builders with various market constraints and ultimate goals for three projects in three climate zones (CZs): Savannah Gardens in Savannah, Georgia (CZ 2); JMC Patrick Square in Clemson, South Carolina (CZ 3); and LaFayette in LaFayette, Georgia (CZ 4). This report documents the design process, computational energy modeling, construction, envelope performance metrics, long-term monitoring results, and successes and failures of the design and execution of these highperformance homes. The JMC Patrick Square project is a single floor with 1,828 ft2 of conditioned living space, three bedrooms, two bathrooms, and an attached two-car garage. This small-scale production builder wanted to increase its level of energy efficiency beyond its current green building practices, including bringing ducts into conditioned space. The team met this goal through a combination of upgrade measures and achieved many Zero Energy Ready Home program requirements. Monitoring the four ducted HPWHs in LaFayette and one in Savannah revealed that HPWH exhaust air impacts attic air during HPWH runtime only, and attic conditions return to previous levels shortly after the HPWH turns off. The HPWH did not appear to impact the loads on the heating and cooling systems, which were also placed in the attic. HPWHs should not be considered dehumidifiers if one is needed in an attic or basement/crawlspace. Ducting the HPWHs did not negatively impact performance compared to other published data of field performance. Changing duct configurations also did not alter the coefficient of performance. HPWHs in efficiency mode (heat pump only) could satisfy hot water demand for most residents. This mode maximizes energy efficiency. Adding .5 in. of insulated sheathing using the Huber ZIP System R Sheathing reduced peak summer temperatures and increased minimum winter temperatures inside the wall assemblies compared to the neighboring home. The neighboring home experienced significantly more risk of condensation and failed the American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc., Standard 160-2009: Criteria for Moisture-Control Design Analysis in Buildings. Despite the fact that energy modeling predicted only a 2% annual savings from the insulated sheathing, preliminary data indicate that reduced heating, ventilating, and air-conditioning runtimes and energy consumption attributed to this measure provide significantly greater savings. Additional research is necessary.

Array↗

An evaluation of circumferentially segmented duct liners

An evaluation of circumferentially segmented duct linings are performed by comparing attenuations of optimized segmented and uniform linings for a range of frequencies and source structures. Broadband suppression and the effects of changes to input modal structure are explored in these comparisons. Credence in the theoretical model was obtained by comparing predictions to results of a carefully controlled experiment performed on the Spinning Mode Synthesizer facility at the NASA Langley Research Center. Excellent agreement was obtained in these comparisons. Predicted results show that the optimized segmented liner is more effective than the optimized uniform liner for the higher order spinning mode sources with a hard-wall/soft-wall admittance variation representing an optimum configuration for the segmented liner. Predicted results also indicate that the segmented linear gives better broadband performance and that its suppression characteristics are not particularly sensitive to changes in input modal structure when compared to that of the uniform lining. It is shown that the greatest suppression benefit of segmented linings occur near mode cuton frequencies and that these benefits are obtained despite reductions in the total amount of acoustic treatment of fifty percent or more.

Watson, W. R.↗

Robot trajectory tracking with self-tuning predicted control

A controller that combines self-tuning prediction and control is proposed for robot trajectory tracking. The controller has two feedback loops: one is used to minimize the prediction error, and the other is designed to make the system output track the set point input. Because the velocity and position along the desired trajectory are given and the future output of the system is predictable, a feedforward loop can be designed for robot trajectory tracking with self-tuning predicted control (STPC). Parameters are estimated online to account for the model uncertainty and the time-varying property of the system. The authors describe the principle of STPC, analyze the system performance, and discuss the simplification of the robot dynamic equations. To demonstrate its utility and power, the controller is simulated for a Stanford arm.

Cui, Xianzhong↗

Effects of elevated temperature on the viscoplastic modeling of graphite/polymeric composites

To support the development of new materials for the design of next generation supersonic transports, a research program is underway at NASA to assess the long term durability of advanced polymer matrix composites (PMC's). One of main objectives of the program was to explore the effects of elevated temperature (23 to 200 C) on the constitutive model's material parameters. To achieve this goal, test data on the observed nonlinear, stress-strain behavior of IM7/5260 and IM7/8320 composites under tension and compression loading were collected and correlated against temperature. These tests, conducted under isothermal conditions using variable strain rates, included such phenomena as stress relaxation and short term creep. The second major goal was the verification of the model by comparison of analytical predictions and test results for off axis and angle ply laminates. Correlation between test and predicted behavior was performed for specimens of both material systems over a range of temperatures. Results indicated that the model provided reasonable predictions of material behavior in load or strain controlled tests. Periods of loading, unloading, stress relaxation, and creep were accounted for.

Gates, Thomas S.↗

Controllability and observability measures for Craig-Bampton substructure representations

Measures of controllability and observability are derived for a Craig-Bampton substructure representation for which fixed interface modes and constraint modes are computed relative to control actuator locations. The measures facilitate modal ordering and act as an index of dynamic completeness for modal truncation. Example problems are given to demonstrate that the controllable and observable reduced substructure models have a greater accuracy in predicting closed loop pole locations, transfer functions and actuation commands than conventional normal mode models.

Triller, Michael J.↗

Fabric geometry distortion during composites processing

Waviness and tow misalignment are often cited as possible causes of data scatter and lower compression stiffness and strength in textile composites. Strength differences of as much as 40 percent have been seen in composites that appear to have the same basic material and structural properties -- i.e., yarn orientation, yarn size, interlacing geometry. Fabric geometry distortion has been suggested as a possible reason for this discrepancy, but little quantitative data or substantial evidence exists. The focus of this research is to contribute to the present understanding of the causes and effects of geometric distortion in textile composites. The initial part of the study was an attempt to gather qualitative information on a variety of textile structures. Existing and new samples confirmed that structures with a significant direction presence would be more susceptible to distortion due to the compaction process. Thus, uniweaves (fiber vol frac: 54-72 percent) biaxial braids (vf: 34-58 percent) demonstrated very little fabric geometry distortion. In stitched panels, only slight buckling of z-direction stitches was observed, primarily near the surface. In contrast, for structures with high compaction ratios -- e.g., large cylindrical yarns (2.5:1) orpowder towpreg (4:1) -- there were visible distortions where previously smooth and periodic undulations were transformed to abrupt changes in direction. A controlled study of the effect of forming pressure on distortion was conducted on type 162 glass plain weave fabrics. Panels (6 x 6 in) were produced via a resin infusion type setup, but with an EPON 815 epoxy resin. Pressures ranging from hand layup to 200 psi were used (vf: 34-54 percent). Photomicrographs indicated that at pressures up to 50 psi, large changes in thickness were due primarily to resin squeeze out. At higher pressures, when intimate contact was made between the layers, there was some tow flattening and in-plane shifting to optimize nesting. However, even at 200 psi the period and amplitude of the tow undulation remained constant, suggesting that for this relatively fine fabric, distortions from compaction were not a problem. Because of the interest in using larger tows (to reduce cost) and more complex structures, tests were also run on 2D triaxial glass braid (113 yd/lb at 0, 225 yd/lb at +/- 45). Forming pressures of 20, 50, 200, and 500 psi were used, and short block compression tests were run. The 500 psi specimen had a 10 percent decrease in modulus and an almost 50 percent decrease in strength (vs. 20 psi). Because the total fiber wgt/panel was kept constant, the thickness varied from 0.32 to 0.22 in (49-70 percent vf). Yet, the strength value is clearly below what would be expected, even with the decrease in thickness. Photomicrographs of these samples will be taken to determine if more fabric distortion exists in the 500 psi specimens. Finally, because the ultimate goal is to be able to predict and control distortion in a variety of textile structures, a model compaction test was developed to directly measure the deformation of the tows during compaction. Layers of dry glass fabric were placed in a mold with a clear plexiglass window. The yarn amplitude and period was then calculated using image analysis of the videotaped deformation. Preliminary tests demonstrated the feasibility of this technique for simple fabrics with large tows.

Chen, Julie↗

The Implications of ACTS Technology on the Requirements of Rain Attenuation Modeling for Communication System Specification and Analysis at 30/20 GHz and Beyond

The advent of the use of the Ka-Band for space communications, coupled with the introduction of digital modulation techniques as well as multiple-beam methodology for satellites, has deemed it necessary to reassess the plethora of rain attenuation prediction models in use. The Advanced Communication Technology Satellite (ACTS) Project, undertaken by the National Aeronautics and Space Administration in 1983, offered such challenges to rain attenuation prediction modeling. Up to 1983, no such single modeling formalism existed that could fill such requirements. Not even the work done by the NASA Propagation Experimenters (NAPEX) Group had envisioned such requirements, so no dynamic Ka-Band data existed from which one could draw conclusions. In this paper, the basic rudiments of what has become to be known as the 'ACTS Rain Attenuation Prediction Model' will be presented. The concept of rain fade mitigation control availability will be introduced. A new evaluation is then presented for the performance of satellite communication systems, in particular, those to be operating within the Ka-Band and above, that will necessarily employ some type of dynamic rain fade mitigation procedure.

Manning, Robert M.↗

Design and Testing of a Blended Wing Body With Boundary Layer Ingestion Nacelles at High Reynolds Numbers

A knowledge-based aerodynamic design method coupled with an unstructured grid Navier-Stokes flow solver was used to improve the propulsion/airframe integration for a Blended Wing Body with boundary-layer ingestion nacelles. A new zonal design capability was used that significantly reduced the time required to achieve a successful design for each nacelle and the elevon between them. A wind tunnel model was built with interchangeable parts reflecting the baseline and redesigned configurations and tested in the National Transonic Facility (NTF). Most of the testing was done at the cruise design conditions (Mach number = 0.85, Reynolds number = 75 million). In general, the predicted improvements in forces and moments as well as the changes in wing pressures between the baseline and redesign were confirmed by the wind tunnel results. The effectiveness of elevons between the nacelles was also predicted surprisingly well considering the crudeness in the modeling of the control surfaces in the flow code. A novel flow visualization technique involving pressure sensitive paint in the cryogenic nitrogen environment used in high-Reynolds number testing in the NTF was also investigated.

Campbell, Richard L.↗

Modeling and Classifying Six-Dimensional Trajectories for Teleoperation Under a Time Delay

Within the context of teleoperating the JSC Robonaut humanoid robot under 2-10 second time delays, this paper explores the technical problem of modeling and classifying human motions represented as six-dimensional (position and orientation) trajectories. A dual path research agenda is reviewed which explored both deterministic approaches and stochastic approaches using Hidden Markov Models. Finally, recent results are shown from a new model which represents the fusion of these two research paths. Questions are also raised about the possibility of automatically generating autonomous actions by reusing the same predictive models of human behavior to be the source of autonomous control. This approach changes the role of teleoperation from being a stand-in for autonomy into the first data collection step for developing generative models capable of autonomous control of the robot.

SunSpiral, Vytas↗

Neural node network and model, and method of teaching same

The present invention is a fully connected feed forward network that includes at least one hidden layer 16. The hidden layer 16 includes nodes 20 in which the output of the node is fed back to that node as an input with a unit delay produced by a delay device 24 occurring in the feedback path 22 (local feedback). Each node within each layer also receives a delayed output (crosstalk) produced by a delay unit 36 from all the other nodes within the same layer 16. The node performs a transfer function operation based on the inputs from the previous layer and the delayed outputs. The network can be implemented as analog or digital or within a general purpose processor. Two teaching methods can be used: (1) back propagation of weight calculation that includes the local feedback and the crosstalk or (2) more preferably a feed forward gradient decent which immediately follows the output computations and which also includes the local feedback and the crosstalk. Subsequent to the gradient propagation, the weights can be normalized, thereby preventing convergence to a local optimum. Education of the network can be incremental both on and off-line. An educated network is suitable for modeling and controlling dynamic nonlinear systems and time series systems and predicting the outputs as well as hidden states and parameters. The educated network can also be further educated during on-line processing.

Parlos, Alexander G.↗