Searcharxiv⌕ Search

arXiv subjects

David Gordon

Publications and source records attributed to David Gordon.

At least 19 recordsLinked to original sources
arXiv2026

The Celestial Reference Frame at K Band: The CRF-K-2025 Catalog

We present an updated K band (24 GHz) celestial reference frame (CRF) constructed from 3.5 million Very Long Baseline Interferometry (VLBI) observations collected during 211 observing epochs between May 2002 and December 2025 using the Very Long Baseline Array (VLBA), the HARTRAO-HOBART26 baseline, the HARTRAO-YEBES40M baseline, and the Korean VLBI Network (KVN) augmented with several other VLBI stations. We have successfully observed and determined precise angular coordinates for 1317 compact extragalactic radio sources, essentially quasars, covering the full sky. This updated K band catalog is designated as CRF-K-2025. The precision of CRF-K-2025 is characterized by median scaled uncertainties of 60 and 104 micro-arc-seconds in right ascension and declination, respectively. The increase in number of observations and sensitivity over earlier K band campaigns has resulted in a catalog with 493 additional sources and a precision approximately 25% better than the ICRF3-K catalog, and similar to the ICRF3-SX catalog. At K band, these quasar radio sources generally show less extended emission than at lower frequencies and thus can potentially provide a more stable long term celestial reference frame than at the standard S/X (2.3/8.4 GHz) observing bands of ICRF3-SX.

astro-ph.GA↗
arXiv2025

AKIBoards: A Structure-Following Multiagent System for Predicting Acute Kidney Injury

Diagnostic reasoning entails a physician's local (mental) model based on an assumed or known shared perspective (global model) to explain patient observations with evidence assigned towards a clinical assessment. But in several (complex) medical situations, multiple experts work together as a team to optimize health evaluation and decision-making by leveraging different perspectives. Such consensus-driven reasoning reflects individual knowledge contributing toward a broader perspective on the patient. In this light, we introduce STRUCture-following for Multiagent Systems (STRUC-MAS), a framework automating the learning of these global models and their incorporation as prior beliefs for agents in multiagent systems (MAS) to follow. We demonstrate proof of concept with a prosocial MAS application for predicting acute kidney injuries (AKIs). In this case, we found that incorporating a global structure enabled multiple agents to achieve better performance (average precision, AP) in predicting AKI 48 hours before onset (structure-following-fine-tuned, SF-FT, AP=0.195; SF-FT-retrieval-augmented generation, SF-FT-RAG, AP=0.194) vs. baseline (non-structure-following-FT, NSF-FT, AP=0.141; NSF-FT-RAG, AP=0.180) for balanced precision-weighted-recall-weighted voting. Markedly, SF-FT agents with higher recall scores reported lower confidence levels in the initial round on true positive and false negative cases. But after explicit interactions, their confidence in their decisions increased (suggesting reinforced belief). In contrast, the SF-FT agent with the lowest recall decreased its confidence in true positive and false negative cases (suggesting a new belief). This approach suggests that learning and leveraging global structures in MAS is necessary prior to achieving competitive classification and diagnostic reasoning performance.

cs.MA↗
arXiv2025

Hybrid Reinforcement Learning and Model Predictive Control for Adaptive Control of Hydrogen-Diesel Dual-Fuel Combustion

Reinforcement Learning (RL) and Machine Learning Integrated Model Predictive Control (ML-MPC) are promising approaches for optimizing hydrogen-diesel dual-fuel engine control, as they can effectively control multiple-input multiple-output systems and nonlinear processes. ML-MPC is advantageous for providing safe and optimal controls, ensuring the engine operates within predefined safety limits. In contrast, RL is distinguished by its adaptability to changing conditions through its learning-based approach. However, the practical implementation of either method alone poses challenges. RL requires high variance in control inputs during early learning phases, which can pose risks to the system by potentially executing unsafe actions, leading to mechanical damage. Conversely, ML-MPC relies on an accurate system model to generate optimal control inputs and has limited adaptability to system drifts, such as injector aging, which naturally occur in engine applications. To address these limitations, this study proposes a hybrid RL and ML-MPC approach that uses an ML-MPC framework while incorporating an RL agent to dynamically adjust the ML-MPC load tracking reference in response to changes in the environment. At the same time, the ML-MPC ensures that actions stay safe throughout the RL agent's exploration. To evaluate the effectiveness of this approach, fuel pressure is deliberately varied to introduce a model-plant mismatch between the ML-MPC and the engine test bench. The result of this mismatch is a root mean square error (RMSE) in indicated mean effective pressure of 0.57 bar when running the ML-MPC. The experimental results demonstrate that RL successfully adapts to changing boundary conditions by altering the tracking reference while ML-MPC ensures safe control inputs. The quantitative improvement in load tracking by implementing RL is an RSME of 0.44 bar.

cs.LG↗
arXiv2025

Incorporating a Deep Neural Network into Moving Horizon Estimation for Embedded Thermal Torque Derating of an Electric Machine

This study presents a novel state estimation approach integrating Deep Neural Networks (DNNs) into Moving Horizon Estimation (MHE). This is a shift from using traditional physics-based models within MHE towards data-driven techniques. Specifically, a Long Short-Term Memory (LSTM)-based DNN is trained using synthetic data derived from a high-fidelity thermal model of a Permanent Magnet Synchronous Machine (PMSM), applied within a thermal derating torque control strategy for battery electric vehicles. The trained DNN is directly embedded within an MHE formulation, forming a discrete-time nonlinear optimal control problem (OCP) solved via the acados optimization framework. Model-in-the-Loop simulations demonstrate accurate temperature estimation even under noisy sensor conditions and simulated sensor failures. Real-time implementation on embedded hardware confirms practical feasibility, achieving computational performance exceeding real-time requirements threefold. By integrating the learned LSTM-based dynamics directly into MHE, this work achieves state estimation accuracy, robustness, and adaptability while reducing modeling efforts and complexity. Overall, the results highlight the effectiveness of combining model-based and data-driven methods in safety-critical automotive control systems.

eess.SY↗
arXiv2025

Safe Reinforcement Learning-based Control for Hydrogen Diesel Dual-Fuel Engines

The urgent energy transition requirements towards a sustainable future stretch across various industries and are a significant challenge facing humanity. Hydrogen promises a clean, carbon-free future, with the opportunity to integrate with existing solutions in the transportation sector. However, adding hydrogen to existing technologies such as diesel engines requires additional modeling effort. Reinforcement Learning (RL) enables interactive data-driven learning that eliminates the need for mathematical modeling. The algorithms, however, may not be real-time capable and need large amounts of data to work in practice. This paper presents a novel approach which uses offline model learning with RL to demonstrate safe control of a 4.5 L Hydrogen Diesel Dual-Fuel (H2DF) engine. The controllers are demonstrated to be constraint compliant and can leverage a novel state-augmentation approach for sample-efficient learning. The offline policy is subsequently experimentally validated on the real engine where the control algorithm is executed on a Raspberry Pi controller and requires 6 times less computation time compared to online Model Predictive Control (MPC) optimization.

eess.SY↗
arXiv2024

Partial Alignment of Astrometric Position Excursions of International Celestial Reference Frame Quasars with Radio Jet Structures

Published analyses of very long baseline interferometry (VLBI) data for the sources included in the third International Celestial Reference Frame (ICRF3) catalog have revealed object-specific, excess astrometric variability and quasi-coherent trajectories as functions of time. A fraction of these sources show markedly elongated distributions of positions on the sky measured with diurnal observations. Here we apply a novel statistical and data-processing method to the diurnal position measurements stretching over 40 years to quantify the degree of elongation and its position angle, for each source with more than 200 data points. We find that 49\% of the examined sources have distribution elongations in excess of 1.3. Robust uncertainties of the directions of maximal astrometric dispersion are computed by the bootstrapping method, and the results are compared with a larger catalog of radio jet directions by Plavin et al. 2022. Nearly one-half of the sources with smaller position angle uncertainties are found to have astrometric position excursions from their mean positions aligned with the radio jet structures within $\pm 30\degr$.

astro-ph.GA↗
arXiv2024

Testing Causal Explanations: A Case Study for Understanding the Effect of Interventions on Chronic Kidney Disease

Randomized controlled trials (RCTs) are the standard for evaluating the effectiveness of clinical interventions. To address the limitations of RCTs on real-world populations, we developed a methodology that uses a large observational electronic health record (EHR) dataset. Principles of regression discontinuity (rd) were used to derive randomized data subsets to test expert-driven interventions using dynamic Bayesian Networks (DBNs) do-operations. This combined method was applied to a chronic kidney disease (CKD) cohort of more than two million individuals and used to understand the associational and causal relationships of CKD variables with respect to a surrogate outcome of >=40% decline in estimated glomerular filtration rate (eGFR). The associational and causal analyses depicted similar findings across DBNs from two independent healthcare systems. The associational analysis showed that the most influential variables were eGFR, urine albumin-to-creatinine ratio, and pulse pressure, whereas the causal analysis showed eGFR as the most influential variable, followed by modifiable factors such as medications that may impact kidney function over time. This methodology demonstrates how real-world EHR data can be used to provide population-level insights to inform improved healthcare delivery.

cs.LG↗
arXiv2024

Metrics of Astrometric Variability in the International Celestial Reference Frame: I. Statistical analysis and selection of the most variable sources

Using very long baseline interferometry data for the sources that comprise the third International Celestial Reference Frame (ICRF3), we examine the quality of the formal source position uncertainties of ICRF3 by determining the excess astrometric variability (unexplained variance) for each source as a function of time. We also quantify multiple qualitatively distinct aspects of astrometric variability seen in the data, using a variety of metrics. Average position offsets, statistical dispersion measures, and coherent trends over time as explored by smoothing the data are combined to characterize the most and least positionally stable ICRF3 sources. We find a notable dependence of the excess variance and statistical variability measures on declination, as is expected for unmodeled ionospheric delay errors and the northern hemisphere dominated network geometries of most astrometric and geodetic observing campaigns.

astro-ph.IM↗
arXiv2024

Robust 1-norm periodograms for analysis of noisy non-Gaussian time series with irregular cadences: Application to VLBI astrometry of quasars

Astronomical time series often have non-uniform sampling in time, or irregular cadences, with long gaps separating clusters of observations. Some of these data sets are also explicitly non-Gaussian with respect to the expected model fit, or the simple mean. The standard Lomb-Scargle periodogram is based on the least squares solution for a set of test periods and, therefore, is easily corrupted by a subset of statistical outliers or an intrinsically non-Gaussian population. It can produce completely misleading results for heavy-tailed distribution of residuals. We propose a robust 1-norm periodogram technique, which is based on the principles of robust statistical estimation. This technique can be implemented in weighted or unweighted options. The method is described in detail and compared with the classical least squares periodogram on a set of astrometric VLBI measurements of the ICRF quasar IERS B0642+449. It is uniformly applied to a collection of 259 ICRF3 quasars each with more than 200 epoch VLBI measurements, resulting in a list of 49 objects with quasi-periodic position changes above the $3\sigma$ level, which warrant further investigation.

astro-ph.IM↗
arXiv2023

The K-band (24 GHz) Celestial Reference Frame determined from Very Long Baseline Interferometry sessions conducted over the past 20 years

The third realization of the International Celestial Reference Frame (ICRF3) was adopted in August 2018 and includes positions of extragalactic objects at three frequencies: 8.4 GHz, 24 GHz, and 32 GHz. In this paper, we present celestial reference frames estimated from Very Long Baseline Interferometry measurements at K-band (24 GHz) including data until June 2022. The data set starts in May 2002 and currently consists of more than 120 24h observing sessions performed over the past 20 years. Since the publication of ICRF3, the additional observations of the sources during the last four years allow maintenance of the celestial reference frame and more than 200 additional radio sources ensure an expansion of the frame. A study of the presented solutions is carried out helping us to understand systematic differences between the astrometric catalogs and moving us towards a better next ICRF solution. We compare K-band solutions (VIE-K-2022b and USNO-K-2022July05) computed by two analysts with two independent software packages (VieVS and Calc/Solve) and describe the differences in the solution strategy. We assess the systematic differences using vector spherical harmonics and describe the reasons for the most prominent ones.

astro-ph.IM↗
arXiv2023

On More than Two Decades of Celestial Reference Frame VLBI Observations in the Deep South: IVS-CRDS (1995-2021)

The International VLBI Service for Geodesy & Astrometry (IVS) regularly provides high-quality data to produce Earth Orientation Parameters (EOP), and for the maintenance and realization of the International Terrestrial and Celestial Reference Frames, ITRF and ICRF. The first iteration of the celestial reference frame (CRF) at radio wavelengths, the ICRF1, was adopted by the International Astronomical Union (IAU) in 1997 to replace the FK5 optical frame. Soon after, the IVS began official operations and in 2009 there was a significant increase in data sufficient to warrant a second iteration of the CRF, ICRF2. The most recent ICRF3, was adopted by the IAU in 2018. However, due to the geographic distribution of observing stations being concentrated in the Northern hemisphere, CRFs are generally weaker in the South due to there being fewer Southern Hemisphere observations. To increase the Southern Hemisphere observations, and the density, precision of the sources, a series of deep South observing sessions was initiated in 1995. This initiative in 2004 became the IVS Celestial Reference Frame Deep South (IVS-CRDS) observing program. This paper covers the evolution of the CRDS observing program for the period 1995 to 2021, details the data products and results, and concludes with a summary of upcoming improvements to this ongoing project.

astro-ph.EP↗
arXiv2023

Automated Dynamic Bayesian Networks for Predicting Acute Kidney Injury Before Onset

Several algorithms for learning the structure of dynamic Bayesian networks (DBNs) require an a priori ordering of variables, which influences the determined graph topology. However, it is often unclear how to determine this order if feature importance is unknown, especially as an exhaustive search is usually impractical. In this paper, we introduce Ranking Approaches for Unknown Structures (RAUS), an automated framework to systematically inform variable ordering and learn networks end-to-end. RAUS leverages existing statistical methods (Cramers V, chi-squared test, and information gain) to compare variable ordering, resultant generated network topologies, and DBN performance. RAUS enables end-users with limited DBN expertise to implement models via command line interface. We evaluate RAUS on the task of predicting impending acute kidney injury (AKI) from inpatient clinical laboratory data. Longitudinal observations from 67,460 patients were collected from our electronic health record (EHR) and Kidney Disease Improving Global Outcomes (KDIGO) criteria were then applied to define AKI events. RAUS learns multiple DBNs simultaneously to predict a future AKI event at different time points (i.e., 24-, 48-, 72-hours in advance of AKI). We also compared the results of the learned AKI prediction models and variable orderings to baseline techniques (logistic regression, random forests, and extreme gradient boosting). The DBNs generated by RAUS achieved 73-83% area under the receiver operating characteristic curve (AUCROC) within 24-hours before AKI; and 71-79% AUCROC within 48-hours before AKI of any stage in a 7-day observation window. Insights from this automated framework can help efficiently implement and interpret DBNs for clinical decision support. The source code for RAUS is available in GitHub at https://github.com/dgrdn08/RAUS .

cs.LG↗
arXiv2022

Position and Proper Motion of Sagittarius A* in the ICRF3 Frame from VLBI Absolute Astrometry

Sagittarius A* (Sgr A*) is a strong, compact radio source believed to be powered by a super-massive black hole at the galactic center. Extinction by dust and gas in the galactic plane prevents observing it optically, but its position and proper motion have previously been estimated using radio interferometry. We present new VLBI absolute astrometry measurements of its precise position and proper motion in the frame of the third realization of the International Celestial Reference Frame, ICRF3. The observations used were made at 52 epochs on the VLBA at K-band (24 GHz) between June 2006 and August 2022. We find the proper motion of Sgr A* to be -3.128 $\pm$ 0.042 mas/yr in right ascension and -5.584 $\pm$ 0.075 mas/yr in declination, or 6.400 $\pm$ 0.073 mas/yr at a position angle of 209.26 $\pm$ 0.51 degrees. We also find its J2000 ICRF3 coordinates at the 2015.0 proper motion epoch to be 17$^h$45$^m$40.034047$^s$ $\pm$ 0.000018$^s$, -29$^o$00'28.21601'' $\pm$ 0.00044''. In galactic coordinates, Sgr A* shows proper motion of -6.396 $\pm$ 0.071 mas/yr in galactic longitude and -0.239 $\pm$ 0.045 mas/yr in galactic latitude, indicating solar motion of 248.0 $\pm$ 2.8 km/sec in the galactic plane and 9.3 $\pm$ 1.9 km/sec towards the north galactic pole.

astro-ph.GA↗
arXiv2022

Machine Learning Integrated with Model Predictive Control for Imitative Optimal Control of Compression Ignition Engines

The high thermal efficiency and reliability of the compression-ignition engine makes it the first choice for many applications. For this to continue, a reduction of the pollutant emissions is needed. One solution is the use of machine learning (ML) and model predictive control (MPC) to minimize emissions and fuel consumption, without adding substantial computational cost to the engine controller. ML is developed in this paper for both modeling engine performance and emissions and for imitating the behaviour of an Linear Parameter Varying (LPV) MPC. Using a support vector machine-based linear parameter varying model of the engine performance and emissions, a model predictive controller is implemented for a 4.5 Cummins diesel engine. This online optimized MPC solution offers advantages in minimizing the \nox~emissions and fuel consumption compared to the baseline feedforward production controller. To reduce the computational cost of this MPC, a deep learning scheme is designed to mimic the behavior of the developed controller. The performance in reducing NOx emissions at a constant load by the imitative controller is similar to that of the online optimized MPC compared to the Cummins production controller. In addition, the imitative controller requires 50 times less computation time compared to that of the online MPC optimization.

eess.SY↗
arXiv2022

Deep Learning based Model Predictive Control for Compression Ignition Engines

Machine learning (ML) and a nonlinear model predictive controller (NMPC) are used in this paper to minimize the emissions and fuel consumption of a compression ignition engine. In this work machine learning is applied in two methods. In the first application, ML is used to identify a model for implementation in model predictive control optimization problems. In the second application, ML is used as a replacement of the NMPC where the ML controller learns the optimal control action by imitating or mimicking the behavior of the model predictive controller. In this study, a deep recurrent neural network including long-short term memory (LSTM) layers are used to model the emissions and performance of an industrial 4.5 liter 4-cylinder Cummins diesel engine. This model is then used for model predictive controller implementation. Then, a deep learning scheme is deployed to clone the behavior of the developed controller. In the LSTM integration, a novel scheme is used by augmenting hidden and cell states of the network in an NMPC optimization problem. The developed LSTM-NMPC and the imitative NMPC are compared with the Cummins calibrated Engine Control Unit (ECU) model in an experimentally validated engine simulation platform. Results show a significant reduction in Nitrogen Oxides (\nox) emissions and a slight decrease in the injected fuel quantity while maintaining the same load. In addition, the imitative NMPC has a similar performance as the NMPC but with a two orders of magnitude reduction of the computation time.

eess.SY↗
arXiv2019

The Myths of Our Time: Fake News

While the purpose of most fake news is misinformation and political propaganda, our team sees it as a new type of myth that is created by people in the age of internet identities and artificial intelligence. Seeking insights on the fear and desire hidden underneath these modified or generated stories, we use machine learning methods to generate fake articles and present them in the form of an online news blog. This paper aims to share the details of our pipeline and the techniques used for full generation of fake news, from dataset collection to presentation as a media art project on the internet.

cs.LG↗
arXiv2019

The Next Generation Celestial Reference Frame

Astrometry, the measurement of positions and motions of the stars, is one of the oldest disciplines in Astronomy, extending back at least as far as Hipparchus' discovery of the precession of Earth's axes in 190 BCE by comparing his catalog with those of his predecessors. Astrometry is fundamental to Astronomy, and critical to many aspects of Astrophysics and Geodesy. In order to understand our planet's and solar system's context within their surroundings, we must be able to to define, quantify, study, refine, and maintain an inertial frame of reference relative to which all positions and motions can be unambiguously and self-consistently described. It is only by using this inertial reference frame that we are able to disentangle our observations of the motions of celestial objects from our own complex path around our star, and its path through the galaxy, and the local group. Every aspect of each area outlined in the call for scientific frontiers in astronomy in the era of the 2020-2030 timeframe will depend on the quality of the inertial reference frame. In this white paper, we propose support for development of radio Very Long Baseline Interferometry (VLBI) capabilities, including the Next Generation Very Large Array (ngVLA), a radio astronomy observatory that will not only support development of a next generation reference frame of unprecedented accuracy, but that will also serve as a highly capable astronomical instrument in its own right. Much like its predecessors, the Very Long Baseline Array (VLBA) and other VLBI telescopes, the proposed ngVLA will provide the foundation for the next three decades for the fundamental reference frame, benefitting astronomy, astrophysics, and geodesy alike.

astro-ph.IM↗