Input Estimators Research Articles

TEFu-Net: A time-aware late fusion architecture for robust multi-modal ego-motion estimation

Ego-motion estimation plays a critical role in autonomous driving systems by providing accurate and timely information about the vehicle’s position and orientation. To achieve high levels of accuracy and robustness, it is essential to leverage a range of sensor modalities to account for highly dynamic and diverse scenes, and consequent sensor limitations.In this work, we introduce TEFu-Net, a Deep-Learning-based late fusion architecture that combines multiple ego-motion estimates from diverse data modalities, including stereo RGB, LiDAR point clouds and GNSS/IMU measurements. Our approach is non-parametric and scalable, making it adaptable to different sensor set configurations. By leveraging a Long Short-Term Memory (LSTM), TEFu-Net produces reliable and robust spatiotemporal ego-motion estimates. This capability allows it to filter out erroneous input measurements, ensuring the accuracy of the car’s motion calculations over time. Extensive experiments show an average accuracy increase of 63% over TEFu-Net’s input estimators and on par results with the state-of-the-art in real-world driving scenarios. We also demonstrate that our solution can achieve accurate estimates under sensor or input failure. Therefore, TEFu-Net enhances the accuracy and robustness of ego-motion estimation in real-world driving scenarios, particularly in challenging conditions such as cluttered environments, tunnels, dense vegetation, and unstructured scenes. As a result of these enhancements, it bolsters the reliability of autonomous driving functions.

Explainable machine learning models for estimating daily dissolved oxygen concentration of the Tualatin River

ABSTRACT Monitoring the quality of river water is of fundamental importance and needs to be taken into consideration when it comes to the research into the hydrological field. In this context, the concentration of the dissolved oxygen (DO) is one of the most significant indicators of the quality of river water. The current study aimed to estimate the minimum, maximum, and mean DO concentrations (DO min, DO max, DO mean) at a gauging station located on Tualatin River, United States. To that end, four machine learning models, such as support vector regression (SVR), multi-layer perceptron (MLP), random forest (RF), and gradient boosting (GB) were established. Root mean square error (RMSE), mean absolute error (MAE), coefficient of correlation (R), and Nash-Sutcliffe efficiency (NSE) metrics were employed to better assess the accuracies of these models. The modeling results demonstrated that the SVR and MLP surpassed the RF and GB models. Despite this, the SVR was concluded to be the best-performing method when used to estimate the DO min, DO max, and DO mean. The best error statistics in the testing phase were related to the SVR model with full (four) inputs to estimate DO mean concentration (RMSE = 0.663 mg/l, MAE = 0.508 mg/l, R = 0.945, NSE = 0.875). Finally, the explainability of the superior models (i.e. SVR models) was conducted using SHapley Additive exPlanations (SHAP) for the first time to estimate DO concentration. In fact, evaluating the explainability of machine learning models can provide useful information about the impact of each of the input estimators used in the procedure of models development. It was concluded that the specific conductance (SC) and followed by water temperature (WT) could provide the most contributions for estimating the DO min, DO max, and DO mean concentrations.

The dependent coordinates in the linearization of constrained multibody systems: Handling and elimination

In this paper, the handling of the dependent coordinates in the linearization of constrained multibody systems is clarified. To this end, a brief but comprehensive overview of linearization approaches used in the field of Multibody System Dynamics is presented. These procedures are illustrated by using a common notation and classified into four groups, depending on the initial form of the nonlinear equations of motion and the selection of the generalized coordinates (a redundant or a minimal set). The handling of the dependent coordinates in each of the procedures is discussed. The linearized equations of motion of constrained multibody systems are of great importance in several applications, such as linear stability and modal analyses, the design of linear state feedback controllers or the building of state and input estimators, like Kalman filters. When modeling a constrained multibody system, a number of coordinates greater than the number of degrees of freedom of the system is usually used. While some linearization approaches make use of the complete set of coordinates, several procedures are based on a coordinate partition in terms of independent and dependent coordinates to reduce the number of linearized equations of motion. In this latter scenario, the role of the dependent coordinates in the linearization is important and, in a general case, these dependent coordinates cannot be ignored to obtain the correct linearized equations of motion. This aspect, which may seem obvious, is not straightforward and is addressed in this work. The importance of considering the set of dependent coordinates in the linearization is demonstrated with a well-acknowledged bicycle benchmark multibody model and an electric kickscooter multibody model with rear and front suspensions. To this end, the Jacobian matrix and the linear stability results of these case studies are computed by considering different choices of independent and dependent coordinates.

Ensemble of ensembles for fine particulate matter pollution prediction using big data analytics and IoT emission sensors

Purpose The study aims to develop a multilayer high-effective ensemble of ensembles predictive model (stacking ensemble) using several hyperparameter optimized ensemble machine learning (ML) methods (bagging and boosting ensembles) trained with high-volume data points retrieved from Internet of Things (IoT) emission sensors, time-corresponding meteorology and traffic data. Design/methodology/approach For a start, the study experimented big data hypothesis theory by developing sample ensemble predictive models on different data sample sizes and compared their results. Second, it developed a standalone model and several bagging and boosting ensemble models and compared their results. Finally, it used the best performing bagging and boosting predictive models as input estimators to develop a novel multilayer high-effective stacking ensemble predictive model. Findings Results proved data size to be one of the main determinants to ensemble ML predictive power. Second, it proved that, as compared to using a single algorithm, the cumulative result from ensemble ML algorithms is usually always better in terms of predicted accuracy. Finally, it proved stacking ensemble to be a better model for predicting PM2.5 concentration level than bagging and boosting ensemble models. Research limitations/implications A limitation of this study is the trade-off between performance of this novel model and the computational time required to train it. Whether this gap can be closed remains an open research question. As a result, future research should attempt to close this gap. Also, future studies can integrate this novel model to a personal air quality messaging system to inform public of pollution levels and improve public access to air quality forecast. Practical implications The outcome of this study will aid the public to proactively identify highly polluted areas thus potentially reducing pollution-associated/ triggered COVID-19 (and other lung diseases) deaths/ complications/ transmission by encouraging avoidance behavior and support informed decision to lock down by government bodies when integrated into an air pollution monitoring system Originality/value This study fills a gap in literature by providing a justification for selecting appropriate ensemble ML algorithms for PM2.5 concentration level predictive modeling. Second, it contributes to the big data hypothesis theory, which suggests that data size is one of the most important factors of ML predictive capability. Third, it supports the premise that when using ensemble ML algorithms, the cumulative output is usually always better in terms of predicted accuracy than using a single algorithm. Finally developing a novel multilayer high-performant hyperparameter optimized ensemble of ensembles predictive model that can accurately predict PM2.5 concentration levels with improved model interpretability and enhanced generalizability, as well as the provision of a novel databank of historic pollution data from IoT emission sensors that can be purchased for research, consultancy and policymaking.

A Global Tracking Sensorless Adaptive PI-PBC Design for Output Voltage Regulation in a Boost Converter Feeding a DC Microgrid

The problem of the output voltage regulation in a DC-DC boost converter feeding a DC microgrid is addressed in this research via the passivity-based control theory with a proportional–integral action (PI-PBC). Two external input estimators were implemented in conjunction with the proposed controller to make it sensorless and adaptive. The first estimator corresponds to the immersion & invariance (I&I) approach applied to calculate the expected value of the DC load, which is modeled as an unknown DC current. The second estimator is based on the disturbance–observer (DO) approach, which reaches the value of the voltage input. The main advantage of both estimators is that these ensure exponential convergence under steady-state operating conditions, and their parametrization only requires the definition of an integral gain. A comparative analysis with simulations demonstrates that the proposed PI-PBC approach is effective in regulating/controlling the voltage profile in unknown DC loads as compared to the adaptive sliding mode controller. Experimental validations have demonstrated that the proposed PI-PBC approach, in conjunction with the I&I and the DO estimators, allowed regulation of the voltage output profile in the terminals of the DC load with asymptotic stability properties and fast convergence times (1.87 ms) and acceptably overshoots (6.1%) when the voltage input varies its magnitude (from 10 to 12 V and from 10 to 8 V) considering that the DC load changed with a square waveform between 1 and 2 A with 100 Hz.

Simultaneous Estimation of the State, Unknown Input, and Output Disturbance in Discrete-Time Linear Systems

A state observer and unknown input and output disturbance estimators are proposed for discrete-time linear systems corrupted by bounded unknown inputs and output disturbances. Necessary and sufficient conditions for the existence of the state observer and disturbance estimators are given. Relationships with the strong observer of Hautus are investigated. The state, unknown input, and output disturbance estimation errors are guaranteed to be <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\ell _{\infty }$</tex-math></inline-formula> -stable with prescribed performance level. The design of the state observer and disturbance estimators are given in terms of linear matrix inequalities. The proposed estimators can be applied to detect adversarial attacks on the communication channels between the controller and actuators and between the plant sensors and the controller.

Evaluation of multi-asset investment strategies with digital assets

The drastic growth of the cryptocurrencies market capitalization boosts investigation of their diversification benefits in portfolio construction. In this paper with a set of classical and modern measurement tools, we assess the out-of-sample performance of eight portfolio allocation strategies relative to the naive 1/N rule applied to traditional and crypto-assets investment universe. Evaluated strategies include a range from classical Markowitz rule to the recently introduced LIBRO approach (Trimborn et al. in Journal of Financial Econometrics 1–27, 2019). Furthermore, we also compare three extensions for strategies with respect to input estimators applied. The results show that in the presence of alternative assets, such as cryptocurrencies, mean–variance strategies underperform the benchmark portfolio. In contrast, CVaR optimization tends to outperform the benchmark as well as geometric optimization, although we find a strong dependence of the former’s success on trading costs. Furthermore, we find evidence that liquidity-bounded strategies tend to perform very well. Thus, our findings underscore the non-normal distribution of returns and the necessity to control for liquidity constraints at alternative asset markets.

Evaluation of Multi-Asset Investment Strategies with Digital Assets

The drastic growth of the cryptocurrencies market capitalization boosts investigation of their diversification benefits in portfolio construction. In this paper with a set of classical and modern measurement tools, we assess the out-of-sample performance of eight portfolio allocation strategies relative to the naive 1=N rule applied to traditional and crypto-assets investment universe. Evaluated strategies include a range from classical Markowitz rule to the recently introduced LIBRO approach, Trimborn et al. (2019). Furthermore, we also compare three extensions for strategies with respect to input estimators applied. The results show that in the presence of alternative assets, such as cryptocurrencies, Mean-Variance strategies under-perform the benchmark portfolio. In contrast, CVaR optimization tends to outperform the benchmark as well as geometric optimization. Furthermore, we find evidence that liquidity-bounded strategies tend to perform very well. Thus, our findings underscore the non-normal distribution of returns as well as the necessity to control for liquidity constraints at alternative assets markets.

Uncertainty measures for economic accounts

The problem of adjusting large systems of estimated economic or social accounts such that they fulfill known functional relationships can be quite complex. For such complex systems, evaluating the accuracy of the estimates after the adjustment is difficult since these estimates are defined by unadjusted initial estimates, the accounting equations and the adjustment method. In this paper, we consider such accounting systems as a single entity and develop scalar uncertainty measures that are based on the first two moments of the joint distribution of final adjusted estimates. Scalar measures can help to effectively communicate to the users the relevant uncertainty of disseminated macro-economic accounts and can assist the producer in choosing and improving adjustment method and input estimators. The proposed approach is illustrated both analytically and by simulation. Applications to supply and use tables and to time series data are presented.

Predictive Alarm Generation for Chemical Processes with Unknown Disturbance

AbstractA predictive warning generation scheme for chemical processes with unknown disturbances is proposed in this paper. The proposed methodology uses predicted states of a process system evaluated from the open loop process model and disturbance estimates. Alarms are issued for two conditions, during a time delay period and at the steady state. Disturbances are estimated using unknown input estimators. The moving horizon predictor combined with bias correction is used to predict the dynamic state of a process for a time‐delay period. To generate warnings for actuator limitations, steady state gain for disturbances along with input constraints are used to check for a feasible solution for using linear programming. A warning is generated to the operator when a feasible solution does not exist. The proposed methodology is demonstrated in a simulated model for a continuous stirred tank heater system (CSTH). Results show an early detection of an abnormal situation that provides the operator a lead time to react in the case of a disturbance affecting the system.

New Method of Faults Diagnostic based on Neuro-Dynamic Sliding Mode for Flat Nonlinear Systems

This paper addresses the problem of simultaneous actuator, process and sensor Fault Detection and Isolation (FDI) for nonlinear system having flatness properties with the presence of disturbances and which are operating in closed-loop. In particular, the nonlinear system is corrupted with additive actuator, process or sensor faults with simultaneous occurrence. In this case, the residual signals might be sensitive to all of these faults that can appear in the system. The proposed FDI method is based on both input and parameter estimators that are designed in parallel. With the flatness property of such system, the design of these two estimators requires information on the measured outputs and their successive derivatives. To estimate these last one, a new scheme of the 2nd-order dynamic sliding mode differentiator is proposed. Residuals are next defined as the difference between the estimated and expected behavior. In order to isolate the faults, dynamic neural networks technique is employed. Besides, comparative study between this new differentiator and the well-known 2nd-order Levant’s differentiator is provided to show the pros and cons of the proposed FDI method. This latter is validated by the simulation results and is carried out on a three tank system.

Moving horizon least-squares input estimation for linear discrete-time

This paper presents a novel moving horizon least-squares input estimation method for linear discrete-time stochastic systems. For systems with completely unknown initial state and no unstable zeros, some existing work showed that asymptotic input reconstruction is possible in the absence of noises. However, under the same condition but with stochastic noises, most existing input estimators, which are designed to optimally deal with noises, fail to ensure asymptotic unbiasedness. In order to address this limitation for linear discrete-time stochastic systems, we characterize necessary and sufficient conditions for input observability and detectability, and propose a moving horizon least-squares input estimator. Based on the conditions for input observability and detectability, it is proved that our proposed input estimator gives an asymptotically unbiased estimate and has minimal estimation error variance over all linear asymptotically unbiased input estimators. Its effectiveness is illustrated by simulation examples involving aircraft sensor and actuator faults.

Non-asymptotic state estimation for a class of linear time-varying systems with unknown inputs

In this paper, we extend the modulating functions method to estimate the state and the unknown input of a linear time-varying system defined by a linear differential equation. We first estimate the unknown input by taking a truncated Jacobi orthogonal series expansion with unknown coefficients which can be estimated by the modulating functions method. Then, we estimate the state by using extended modulating functions and the estimated input. Both input and state estimators are given by exact integral formulae involving modulating functions and the noisy output. Hence, estimations at different instants can be non-asymptotically obtained using a sliding window of finite length. Numerical results are given to show the accuracy and the robustness of the proposed estimators against corrupting noises.

Less Is More: Mixed-Initiative Model-Predictive Control With Human Inputs

This paper presents a new method for injecting human inputs into mixed-initiative interactions between humans and robots. The method is based on a model-predictive control (MPC) formulation, which inevitably involves predicting the system (robot dynamics as well as human input) into the future. These predictions are complicated by the fact that the human is interacting with the robot, causing the prediction method itself to have an effect on future human inputs. We investigate and develop different prediction schemes, including fixed and variable horizon MPCs and human input estimators of different orders. Through a search-and-rescue-inspired human operator study, we arrive at the conclusion that the simplest prediction methods outperform the more complex ones, i.e., in this particular case, less is indeed more.

The robustness of the hyperbolic efficiency estimator

The robustness properties of a specific type of orientation in the context of efficiency measurement using partial frontiers are investigated. This so called unconditional hyperbolic quantile estimator of efficiency has been recently studied and can be seen as an extension of the input/output methodology of partial frontiers that was introduced previously. The influence function as well as the breakdown point of this fully non-parametric and unconditional estimator are derived for a complete multivariate setup (multiple inputs and outputs). Like for the input and output quantile estimators, the hyperbolic quantile estimator is B-robust but unlike the two former types of estimator its breakdown point does not depend on the actual input or output level of the production unit. Some examples are given to assess the relevance of this type of estimator and to show the differences with the input and output quantile estimators of efficiency from both a robustness and a statistical efficiency point of view. Finally, a real life example is used to illustrate how the hyperbolic efficiency estimator might be used in a robust context.

Rainfall Drop-Size Estimators for Weibull Probability Distribution Using Method of Moments Technique

This paper proposes a new approach for deriving the input estimators for the Weibull probability rain drop-size distribution (DSD) using the Method of Moments (MM). The Stirling's approximation is used to estimate the gamma function which is part of the raw moment function of the Weibull probability distribution. The parameters No, �� and �� are estimated using a comparative analysis with the likely measured moments of a statistical data. The new parameters are then tested with collected data at Durban, South Africa by using least squares regression fitting technique to derive their power-law relationships with rainfall rate. The results show that the proposed Weibull distribution fits well with the measured data at tested rainfall rates. In comparison with the ITU-R P. 838-3, the average RMSE of its specific attenuation is 0.99 for horizontal polarization and 1.40 for vertical polarization at selected rainfall rates for a frequency range of 0 to 100 GHz.

Unknown disturbance inputs estimation based on a state functional observer design

We consider the problem of function of state plus unknown input estimation of a linear time-invariant system using only the measured outputs. Two reduced-order input estimators built upon a state functional observer are proposed. The first is an extension of a state/input estimator, while the second is based on adaptive observer design technique. The proposed estimator can be designed under less restrictive conditions than those of the previous work, and unlike some of the past studies the proposed observer can be designed for certain nonminimum phase systems.

Application of input estimation techniques to charge estimation and control in automotive engines

The performance of control, estimation and diagnostics algorithms of modern automotive engine management systems can be significantly enhanced using input observers. This paper describes several case studies along these lines in the area of charge estimation and control for gasoline and diesel engines, and illustrates the conclusions with experimental results obtained from a vehicle and from an engine dynamometer. A self-contained overview of some of the popular input estimation techniques and their properties is provided. Pointers are given that show how input estimators can be integrated within more complex adaptation and control algorithms using specific automotive applications as examples.

A New Method of Solving Noisy Abel-Type Equations

A new approach to solving noisy integral equations of the first kind is applied to the family of Abel equations. Such equations play a role in stereology (Wicksell's unfolding problem), medicine, engineering, and astronomy. The method is based on an expansion in an arbitrary orthonormal basis, coupled with exact inversion of the integral operator. The inverse appears in the Fourier coefficients of the expansion, where it can be carried over to the usually well-behaved basis elements in the form of the adjoint. This method is an alternative to Tikhonov regularization, regularization of the inverse of the operator itself, or a wavelet-vaguelette/singular-value decomposition. The method is particularly interesting in irregularity of the kernel, the input, or both. Because knowledge of the spectral properties of the operator is not required, the method is also of interest in regular cases where these spectral properties are not sufficiently known or are hard to deal with. For smooth input functions, the simple basis of trigonometric functions yields input estimators whose mean integrated squared error converges at the optimal rate for the entire family of Abel operators. This can be shown when smooth wavelets are used for Abel operators with index smaller than 1/2, and when the Haar wavelet is used for operators with index larger than 1/2.

Optimal and self-tuning white noise estimators with applications to deconvolution and filtering problems

Using the innovation analysis method in the time domain, based on the autoregressive moving average (ARMA) innovation model, this paper presents a unified white noise estimation theory that includes both input and measurement white noise estimators, and presents a new steady-state optimal state estimation theory. Non-recursive optimal state estimators are given, whose recursive version gives a steady-state Kalman filter, where a new algorithm of the Kalman filter gain is proposed. Two new algorithms of the Kalman predictor gain are also derived. Local asymptotic stability of the Kalman filter is proved. The classical Kalman filtering theory is extended and modified. The method used covers unstable systems with correlation noises and singular transition matrix, and also covers the self-tuning white noise, signal and state estimators, when the noise statistics is unknown. To illustrate, three simulation examples are given.