Search

Estimation of unknown noise covariances in a Kalman filter is a problem of significant practical interest in a wide array of applications. This paper presents a single-pass stochastic gradient descent (SGD) algorithm for noise covariance estimation for use in adaptive Kalman filters applied to non-stationary systems where the noise covariances can occasionally jump up or down by an unknown magnitude. Unlike our previous batch method or our multi-pass decision-directed algorithm, the proposed streaming algorithm reads measurement data exactly once and has similar root mean square error (RMSE). The computational efficiency of the new algorithm stems from its one-pass nature, recursive fading memory estimation of the sample cross-correlations of the innovations, and the RMSprop accelerated SGD algorithm. The comparative evaluation of the proposed method on a number of test cases demonstrates its computational efficiency and accuracy.

In this paper, we consider the problem of state-of-charge estimation for rechargeable batteries. Coulomb counting is a well-known method for estimating the state of charge, and it is regarded as accurate as long as the battery capacity and the beginning state of charge are known. The Coulomb counting approach, on the other hand, is prone to inaccuracies from a variety of sources, and the magnitude of these errors has not been explored in the literature. We formally construct and quantify the state-of-charge estimate error during Coulomb counting due to four types of error sources: (1) current measurement error; (2) current integration approximation error; (3) battery capacity uncertainty; and (4) timing oscillator error/drift. It is demonstrated that the state-of-charge error produced can be either time-cumulative or state-of-charge-proportional. Time-cumulative errors accumulate over time and have the potential to render the state-of-charge estimation utterly invalid in the long term.The proportional errors of the state of charge rise with the accumulated state of charge and reach their worst value within one charge/discharge cycle. The study presents methods for reducing time-cumulative and state-of-charge-proportional mistakes through simulation analysis.

Estimation of unknown noise covariances in a Kalman filter is a problem of significant practical interest in a wide array of applications. Although this problem has a long history, reliable algorithms for their estimation were scant, and necessary and sufficient conditions for identifiability of the covariances were in dispute until recently. Necessary and sufficient conditions for covariance estimation and a batch estimation algorithm. This paper presents stochastic gradient descent (SGD) algorithms for noise covariance estimation in adaptive Kalman filters that are an order of magnitude faster than the batch method for similar or better root mean square error (RMSE) and are applicable to non-stationary systems where the noise covariances can occasionally jump up or down by an unknown magnitude. The computational efficiency of the new algorithm stems from adaptive thresholds for convergence, recursive fading memory estimation of the sample cross-correlations of the innovations, and accelerated SGD algorithms. The comparative evaluation of the proposed method on a number of test cases demonstrates its computational efficiency and accuracy.

Reliability assessment of power semiconductor devices requires accelerated stress tests. Different test methods might produce different results. In this paper, two different test methods are used for on-state resistance degradation for Si MOSFETs: High frequency cycling and low frequency cycling. Presented results indicate that the degradation of on-state resistance occurs irrespective of which method is used. Additionally, gate-to-source capacitance degradation is shown where results show that the gate-to-source impedance degrades over time.

During the course of an epidemic, one of the most challenging tasks for authorities is to decide what kind of restrictive measures to introduce and when these should be enforced. In order to take informed decisions in a fully rational manner, the onset of a critical regime, characterized by an exponential growth of the contagion, must be identified as quickly as possible. Providing rigorous quantitative tools to detect such an onset represents an important contribution from the scientific community to proactively support the political decision makers. In this paper, leveraging the quickest detection theory, we propose a mathematical model of the COVID-19 pandemic evolution and develop decision tools to rapidly detect the passage from a controlled regime to a critical one. A new sequential test—referred to as MAST (mean-agnostic sequential test)—is presented, and demonstrated on publicly available COVID-19 infection data from different countries. Then, the performance of MAST is investigated for the second pandemic wave, showing an effective trade-off between average decision delay $$\Delta$$ Δ and risk $$R$$ R , where $$R$$ R is inversely proportional to the time required to declare the need to take unnecessary restrictive measures. To quantify risk, in this paper we adopt as its proxy the average occurrence rate of false alarms, in that a false alarm risks unnecessary social and economic disruption. Ideally, the decision mechanism should react as quick as possible for a given level of risk. We find that all the countries share the same behaviour in terms of quickest detection, specifically the risk scales exponentially with the delay, $$R \sim \exp {(-\omega \Delta )}$$ R ∼ exp ( - ω Δ ) , where $$\omega$$ ω depends on the specific nation. For a reasonably small risk level, say, one possibility in ten thousand (i.e., unmotivated implementation of countermeasures every 27 years, on the average), the proposed algorithm detects the onset of the critical regime with delay between a few days to 3 weeks, much earlier than when the exponential growth becomes evident. Strictly from the quickest-detection perspective adopted in this paper, it turns out that countermeasures against the second epidemic wave have not always been taken in a timely manner. The developed tool can be used to support decisions at different geographic scales (regions, cities, local areas, etc.), levels of risk, instantiations of controlled/critical regime, and is general enough to be applied to different pandemic time-series. Additional analysis and applications of MAST are made available on a dedicated website.

This paper presents a machine learning approach using Neural Networks for the detection and localization of a single streaking target from an optical sensor's Focal Plane Array (FPA). Current state-of-the-art methods utilize model-based probabilistic techniques, such as the maximum likelihood method. The image data for training and the concomitant ground truth for detection and localization is generated via simulation due to lack of sufficient amounts of real-world data. The images were generated assuming that the target's point spread function (PSF) is Gaussian and moves during the FPA's integration time. Also, a noise model for each optical pixel on the FPA is used that is consistent with a Poisson model of the number of non-target originated photons. Our approach divides the problem into two parts: streak detection and localization. To simplify neural network architecture selection and training, we split the large images into smaller sub-images. The noisy sub-images were preprocessed by an autoencoder with Convolutional layers, which preserved the target intensity. Next, we used the denoised sub-images from the autoencoder to detect streaks. For localization, the Neural Network was given the original sub-images to predict the pixel locations of a streak's start and end points. All three processing steps utilized Convolutional layers. The performance of the detection model was evaluated in terms of confusion matrices at different noise intensities, as well as the Receiver Operating Characteristic (ROC) curves at a constant noise intensity. For the localization model, we evaluated the Mean Squared Error (MSE). These results were compared to the Generalized Likelihood Ratio Test (GLRT) for detection and an existing matched filter-based maximum likelihood method for localization. We found that the machine learning models were 340 times faster for detection and approximately 360 times faster for localization on full size 256x256 images. In addition, the detection accuracy of the machine learning model was substantially better than the maximum likelihood method (area under the ROC of 0.94 versus 0.85) and the MSE was significantly lower (0.0261 versus 0.0622 on 32x32 sub-images and 0.0351 versus 0.243 on 256x256 original images). These results are significant for time-critical systems, such as missile defense systems (e.g., C-RAM). Aside from missile defense, the approach presented here can be applied to counter many other weapons that exhibit similar linear streak trajectories, such as torpedoes. In this case, sonar images would be used rather than images from an optical sensor.

Weather-impacted asset routing is a complex problem, involving nonlinear, non-convex costs due to spatiotemporally varying uncertainty of weather parameters over a finite time horizon. To this end, we have augmented a previously published tool for multiobjective planning and asset routing, termed TMPLAR, with a new approximate dynamic programming-based Pareto optimization (NAPO) algorithm. TMPLAR is a mixed-initiative tool to solve the asset routing problem in dynamic and uncertain environments. It is built upon multi-objective dynamic programming algorithms to route assets in a timely fashion, while considering objectives, such as fuel efficiency, voyage time, distance, and adherence to real world constraints (asset vehicle limits, navigator-specified deadlines, etc.). The asset routing problem is exacerbated by the need to address multiple conflicting objectives, spatial and temporal uncertainty associated with the weather and multiple constraints on asset operation. The NAPO algorithm optimizes weather-based objectives in a reasonable amount of time, optimizing arrival and departure times at waypoints, asset speed and bearing. The key algorithmic contribution is a fast approximate method for substantially containing the computational complexity by generating the Pareto-front of the multi-objective shortest path problem for networks with stochastic non-convex edge costs, utilizing approximate dynamic programming and clustering techniques. The proposed algorithm is validated and we compare its performance with the new approach to multi-objective A* (NAMOA*).

This paper develops an easily-implementable version of Page's CUSUM quickest-detection test, designed to work in certain composite hypothesis scenarios with time-varying data statistics. The decision statistic can be cast in a recursive form and is particularly suited for on-line analysis. By back-testing our approach on publicly-available COVID-19 data we find reliable early warning of infection flare-ups, in fact sufficiently early that the tool may be of use to decision-makers on the timing of restrictive measures that may in the future need to be taken., Accepted to be published in IEEE Signal Processing Letters

Even if path planning can be solved using standard techniques from dynamic programming and control, the problem can also be approached using probabilistic inference. The algorithms that emerge using the latter framework bear some appealing characteristics that qualify the probabilistic approach as a powerful alternative to the more traditional control formulations. The idea of using estimation on stochastic models to solve control problems is not new and the inference approach considered here falls under the rubric of Active Inference (AI) and Control as Inference (CAI). In this work, we look at the specific recursions that arise from various cost functions that, although they may appear similar in scope, bear noticeable differences, at least when applied to typical path planning problems. We start by posing the path planning problem on a probabilistic factor graph, and show how the various algorithms translate into specific message composition rules. We then show how this unified approach, presented both in probability space and in log space, provides a very general framework that includes the Sum-product, the Max-product, Dynamic programming and mixed Reward/Entropy criteria-based algorithms. The framework also expands algorithmic design options for smoother or sharper policy distributions, including generalized Sum/Max-product algorithm, a Smooth Dynamic programming algorithm and modified versions of the Reward/Entropy recursions. We provide a comprehensive table of recursions and a comparison through simulations, first on a synthetic small grid with a single goal with obstacles, and then on a grid extrapolated from a real-world scene with multiple goals and a semantic map.

The COVID-19 pandemic has, worldwide and up to December 2020, caused over 1.7 million deaths, and put the world's most advanced healthcare systems under heavy stress. In many countries, drastic restrictive measures adopted by political authorities, such as national lockdowns, have not prevented the outbreak of new pandemic's waves. In this article, we propose an integrated detection-estimation-forecasting framework that, using publicly available data, is designed to: (i) learn relevant features of the pandemic (e.g., the infection rate); (ii) detect as quickly as possible the onset (or the termination) of an exponential growth of the contagion; and (iii) reliably forecast the pandemic evolution. The proposed solution is validated by analyzing the COVID-19 second and third waves in the USA., Comment: Accepted to be published in IEEE Communications Magazine, feature topic "Networking Technologies to Combat the COVID-19 Pandemic"

In this article, we present the datasets collected from nine different Li-ion batteries. These datasets contain voltage, current and time measurements during a full charge-discharge cycle of a battery at very low current (that is nearly at C / 30 rate). Such low current rate data is suitable for open circuit voltage characterization. The collection of this data was done through the use of an Arbin battery cycler and a thermal chamber was used to control the test temperature. Data were collected over a wide range of temperatures from − 25 ∘ C to 50 ∘ C.

Electric vehicles are set to be the dominant form of transportation in the near future and Lithium-based rechargeable battery packs have been widely adopted in them. Battery packs need to be constantly monitored and managed in order to maintain the safety, efficiency and reliability of the overall electric vehicle system. A battery management system consists of a battery fuel gauge, optimal charging algorithm, and cell/thermal balancing circuitry. It uses three non-invasive measurements from the battery, voltage, current and temperature, in order to estimate crucial states and parameters of the battery system, such as battery impedance, battery capacity, state of charge, state of health, power fade, and remaining useful life. These estimates are important for the proper functioning of optimal charging algorithms, charge and thermal balancing strategies, and battery safety mechanisms. Approach to robust battery management consists of accurate characterization, robust estimation of battery states and parameters, and optimal battery control strategies. This paper describes some recent approaches developed by the authors towards developing a robust battery management system.

A novel decision support system comprising of several interference identification algorithms for the detection of hazardous (i.e., risk to life, assets, and/or mission) voyages and water space allocations in maritime operations is proposed. Given a specified spatio-temporal region for each polygonal or ellipsoidal object, the objective is to identify interference, defined as overlap, among the assigned regions. The challenge is to fuse information on bathymetry, obstacles, and planned spatial assignments to identify conflicts among any combination of polygons or ellipses. We briefly review data structures for the representation of spatio-temporal regions computationally, and efficient algorithms for collision detection. Specifically, we propose a two-phase solution approach to interference identification using time-parameterized R-trees, linear programming, and quadratic programming. The proposed interference detection algorithms can detect overlaps among thousands of polygons and elliptic regions. The resulting decision support tool, conflict identification, is part of a larger system for water space planning.

Electric power steering (EPS) is a fundamental part of an automotive system. Deviations from its anticipated operation have an impact on vehicle driving performance and handling, and could cause severe safety concerns. Reliable and efficient fault detection and diagnosis (FDD) methods are needed for EPS to guarantee the safe operation of the vehicle and for improved repairability of faulty components. In this paper, we develop an integrated model-based and data-driven FDD approach for the EPS system. Specifically, we develop a physics-based model of the EPS system and conduct a number of fault injection experiments to derive fault-sensor measurement dependencies. Then, we investigate various FDD schemes to detect and isolate the faults with special emphasis on rough set-theory-based fault classification. Finally, we compare its fault-classification accuracies with those from traditional classification methods. We demonstrate that the rough set-theory-based FDD approach is robust to missing data.

A battery-management system (BMS) is essential for the safe, reliable, and efficient operation of a battery pack. The BMS uses three noninvasive measurements from the battery-voltage, current, and temperature-to gauge a battery's state of charge (SOC) and state of health (SOH). These estimates are used in BMS functions, such as the generation of optimal charging waveforms, cell balancing, and the activation of safety mechanisms.