Search

Non-Rigid Structure-from-Motion (NRSfM) reconstructs a deformable 3D object from the correspondences established between monocular 2D images. Current NRSfM methods lack statistical robustness, which is the ability to cope with correspondence errors.This prevents one to use automatically established correspondences, which are prone to errors, thereby strongly limiting the scope of NRSfM. We propose a three-step automatic pipeline to solve NRSfM robustly by exploiting isometry. Step 1 computes the optical flow from correspondences, step 2 reconstructs each 3D point's normal vector using multiple reference images and integrates them to form surfaces with the best reference and step 3 rejects the 3D points that break isometry in their local neighborhood. Importantly, each step is designed to discard or flag erroneous correspondences. Our contributions include the robustification of optical flow by warp estimation, new fast analytic solutions to local normal reconstruction and their robustification, and a new scale-independent measure of 3D local isometric coherence. Experimental results show that our robust NRSfM method consistently outperforms existing methods on both synthetic and real datasets., Accepted in TPAMI 2021

Due to the simultaneous existence of model uncertainties and external disturbances, designing automated ground vehicle path-following controllers is recognized as a challenging task. The $H_{∞}$ robust control methodology, as one of the accomplished strategies for controller robustification, has been commonly adopted by researchers to address the vehicle path-tracking problems. Nevertheless, despite its advantages, the $H_{∞}$ controller is only capable of limiting the total ``energy'' of the tracking errors. On the other hand, from a safety standpoint, constraining the ``peak'' of the tracking errors may carry an equal or more importance. To establish a guaranteed upper bound on the path-tracking errors, this paper proposes a novel methodology to synthesis the ground vehicle path-following controller in light of the energy-to-peak robust control theory. Additionally, to address the time-varying uncertainties presented in the tire dynamics, robust stabilization constraints based upon the small-gain theorem are also formulated into the overall controller design problem. Comparative study regarding the disturbance rejection performance between the proposed controller and the conventional $H_{∞}$ approach is conducted via CarSim-Simulink joint simulations. Furthermore, the robustness and disturbance attenuation ability of the energy-to-peak path-tracking controller is experimentally verified on a scaled car.

In this letter, a robust quantized sampled–data controller is provided for a class of nonlinear systems affected by time–varying uncertainties, actuation disturbances and measurement noises. Sufficient conditions based on linear matrix inequalities and ensuring the existence of the proposed robust quantized sampled–data controller are given. Quantization of both state measurements and input signals is simultaneously considered. Input–to–state stability redesign technique is used in order to attenuate the effects of bounded actuation disturbances and of bounded observation errors. It is proved that, under suitably fast sampling and accurate quantization of the input/output channels, the proposed controller achieves the semi–global practical stability, with arbitrarily small final target ball, of the related quantized sampled–data closed–loop system provided that the observation errors do not affect (or affect marginally) the robustification term added in the controller and, that the bounds of the actuation disturbances as well as of the observation errors are a priori known. The theory here developed includes also the cases of time–varying sampling intervals and of non–uniform quantization of the input/output channels as well as the stability analysis of the inter–sampling system behavior. The provided results are validated through an example of one–link manipulator.

Spatial units typically vary over many of their characteristics, introducing potential unobserved heterogeneity which invalidates commonly used homoskedasticity conditions. In the presence of unobserved heteroskedasticity, standard methods based on the (quasi-)likelihood function generally produce inconsistent estimates of both the spatial parameter and the coefficients of the exogenous regressors. A robust generalized method of moments estimator as well as a modified likelihood method have been proposed in the literature to address this issue. The present paper constructs an alternative indirect inference approach which relies on a simple ordinary least squares procedure as its starting point. Heteroskedasticity is accommodated by utilizing a new version of continuous updating that is applied within the indirect inference procedure to take account of the parametrization of the variance-covariance matrix of the disturbances. Finite sample performance of the new estimator is assessed in a Monte Carlo study and found to offer advantages over existing methods. The approach is implemented in an empirical application to house price data in the Boston area, where it is found that spatial effects in house price determination are much more significant under robustification to heterogeneity in the equation errors.

A successful index helps policy decision makers identify benchmark performances and trends and set policy priorities. Selecting representative variables from a large number of potential candidates in the system is crucial to the success of index construction. A robust and sparse method reflects an urgent need to effectively select useful variables for index construction. Although the least absolute shrinkage and selection operator (Lasso) is a popular technique for variable selection, it suffers from the influence of outlier or noise. In this paper, we propose a robust Lasso with a generic insensitive and adaptive loss function (GIA-Lasso) for variable selection. The generic loss function can achieve great robustness against outliers by adjusting an insensitive parameter, an elastic interval parameter, and an adaptive robustification parameter. The $$L_1$$ -norm regularization term and the supervised selection process in GIA-Lasso ensure that the most representative variables are selected. The variable selection results of the Financial Conditions Index and Innovation and Entrepreneurship Index confirm that GIA-Lasso is not only robust against outliers, but also selects representative variables. The Granger causality test further proves the reasonability of the selected variables.

We investigate competitive online algorithms for online convex optimization (OCO) problems with linear in-stage costs, switching costs and ramp constraints. While OCO problems have been extensively studied in the literature, there are limited results on the corresponding online solutions that can attain small competitive ratios. We first develop a powerful computational framework that can compute an optimized competitive ratio based on the class of affine policies. Our computational framework can handle a fairly general class of costs and constraints. Compared with other competitive results in the literature, a key feature of our proposed approach is that it can handle scenarios where infeasibility may arise due to hard feasibility constraints. Second, we design a robustification procedure to produce an online algorithm that can attain good performance for both average-case and worst-case inputs. We conduct a case study on Network Functions Virtualization (NFV) orchestration and scaling to demonstrate the effectiveness of our proposed methods.

This article proposes a design approach, based on the integral sliding mode control paradigm, devoted to give robustness to multiagent systems executing arbitrary distributed optimization and consensus protocols which do not take this feature into account. Robustness is understood as the capability of rejecting the effect of exogenous disturbances, parameter uncertainties, and uncertain couplings between the agents dynamics, by achieving the same emerging behavior as that corresponding to a reference multiagent system (MAS) designed to achieve a given coordination objective in the nominal case. The proposed approach yields a distributed state feedback which can seamlessly be integrated into existing distributed optimization and cooperative control protocols which are usually prone to disturbances and uncertainties corrupting the MAS dynamics. Nonsmooth Lyapunov analysis supports the claimed properties. Numerical simulations, showing how popular distributed optimization and consensus protocols can effectively be robustified are discussed.

Given the effect that outliers can have on regression and specification testing, a vastly used robustification strategy by practitioners consists in: (i) starting the empirical analysis with an outlier detection procedure to deselect atypical data values; then (ii) continuing the analysis with the selected non-outlying observations. The repercussions of such robustifying procedure on the asymptotic properties of subsequent inferential procedures are, however, underexplored. We study the effects of such a strategy on testing for heteroscedasticity. Specifically, using weighted and marked empirical processes of residuals theory, we show that the White test implemented after the outlier detection and removal is asymptotically chi-square if the underlying errors are symmetric. In a simulation study, we show that-depending on the type of outliers-the standard White test can be either severely undersized or oversized, as well as have trivial power. The statistic applied after deselecting outliers has good finite sample properties under symmetry but can suffer from size distortions under asymmetric errors. Given these results, we devise an empirical modeling strategy to guide practitioners whose preferred approach is to remove outliers from the sample.

In this paper, we address a class of distributed optimization problems with non-strictly convex cost functions in the presence of communication delays between an agent and a coordinator. To this en...

International audience; In this letter, we consider the robustification of port-Hamiltonian systems with respect to matched time-varying disturbances generated by an exo-system, which is assumed to be known. This letter is an extension, to the case of time-varying disturbances, of recent integral action techniques, which are able to reject constant disturbances only. The approach is then extended to the special case of sinusoidal disturbances with unknown frequency. The main results of this letter are demonstrated on a 2 degree-of-freedom robotic manipulator.

This article addresses a convex optimization problem that suffers from communication delays between a central collector and subsystems. First, we introduce continuous-time optimization dynamics that mirror the alternating direction method of multipliers architecture. Furthermore, we reveal that the proposed optimization dynamics have an architecture similar to that of bilateral teleoperation systems. By virtue of this passivity-based perspective, we apply a scattering transformation, which compensates for communication delays. Finally, we demonstrate the effectiveness of the proposed dynamics through numerical simulations of a grasp force optimization problem.

Recently, the robustification of principal component analysis (PCA) has attracted much research attention in numerous areas of science and engineering. The most popular and successful approach is to model the robust PCA problem as a low-rank matrix recovery problem in the presence of sparse corruption. With this model, the nuclear norm and $\ell _{1}$ -norm penalties are usually used for low-rank and sparsity promotion. Although the nuclear norm and $\ell _{1}$ -norm are favorable due to their convexity, they have a bias problem. In comparison, nonconvex penalties can be expected to yield better recovery performance. In this paper, we consider a formulation for robust PCA using generalized nonconvex penalties for low-rank and sparsity inducing. This nonconvex formulation is efficiently solved by a multi-block alternative direction method of multipliers (ADMM) algorithm. A sufficient condition for the convergence of this new ADMM algorithm has been derived. Furthermore, to address the important issue of nonconvex penalty selection, we have evaluated the new algorithm via numerical experiments in various low-rank and sparsity conditions. The results indicate that, “exact” recovery of the low-rank principle component can be achieved only by nonconvex regularization. MATLAB code is available at https://github.com/FWen/RPCA.git .

This work focuses on the robust optimization of a permanent magnet (PM) synchronous machine while considering a driving cycle. The robustification is obtained by considering uncertainties of different origins. Firstly, there are geometrical uncertainties caused by manufacturing inaccuracies. Secondly, there are uncertainties linked to different driving styles. The final set of uncertainties is linked to ambient parameters such as traffic and weather conditions. The optimization goal is to minimize the PM's volume while maintaining a desired machine performance measured by the energy efficiency over the driving cycle and the machine's maximal torque. The magnetic behavior of the machine is described by a partial differential equation (PDE) and is simulated by the finite-element method employing an affine decomposition to avoid reassembling of the system of equations due to the changing PM geometry. The Sequential Quadratic Programming algorithm is used for the optimization. Stochastic collocation is applied to compute moments of stochastic quantities. The robustness of the optimized configurations is validated by a Monte Carlo sampling. It is found that the uncertainties in driving style and road conditions have significant influence on the optimal PM configuration., 5 pages, 4 figures, 2 tables, Compumag 2019 conference

To increase the contribution of offshore wind energy to the global energy mix in an economically sustainable manner, it is required to reduce the costs associated with the production and operation of offshore wind turbines (OWTs). One of the largest uncertainties and sources of conservatism in design and lifetime prediction for OWTs is the determination of the global damping level of the OWT. Estimation of OWT damping based on field measurement data has hence been subject to considerable research attention and is based on the use of (preferably operational) vibration data obtained from sensors mounted on the structure. As such, it is an output-only problem and can be addressed using state-of-the-art operational modal analysis (OMA) techniques, reviewed in this paper. The evolution of classical time- and frequency-domain OMA techniques has been reviewed; however, the literature shows that the OWT vibration data are often contaminated by rotor speed harmonics of significantly high energy located close to structural modes, which impede classical damping identification. Recent advances in OMA algorithms for known or unknown harmonic frequencies can be used to improve identification in such cases. Further, the transmissibility family of OMA algorithms is purported to be insensitive to harmonics. Based on this review, a classification of OMA algorithms is made according to a set of novel suitability criteria, such that the OMA technique appropriate to the specific OWT vibration measurement setup may be selected. Finally, based on this literature review, it has been identified that the most attractive future path for OWT damping estimation lies in the combination of uncertain non-stationary harmonic frequency measurements with statistical harmonic isolation to enhance classical OMA techniques, orthogonal removal of harmonics from measured vibration signals, and in the robustification of transmissibility-based techniques.

This paper addresses the robust gait control for planar and passive biped robots using approaches based on energy properties. Energy pumping-and-damping passivity-based control is used to increase the robustness against uncertainties on the initial conditions of the passive gait exhibited by planar biped robots. The stability analysis is carried out by exploiting the system’s passivity and the hybrid zero dynamics method. Besides, the proposed approach is applied to new gaits that are generated using interconnection and damping assignment passivity-based control. The performance of the proposed design is evaluated through numerical simulations and compared with an existing technique.

Visual-based recommender systems (VRSs) enhance recommendation performance by integrating users' feedback with the visual features of items' images. Recently, human-imperceptible image perturbations, defined adversarial samples, have been shown capable of altering the VRSs performance, for example, by pushing (promoting) or nuking (demoting) specific categories of products. One of the most effective adversarial defense methods is adversarial training (AT), which enhances the robustness of the model by incorporating adversarial samples into the training process and minimizing an adversarial risk. The AT effectiveness has been verified on defending DNNs in supervised learning tasks such as image classification. However, the extent to which AT can protect deep VRSs, against adversarial perturbation of images remains mostly under-investigated. This work focuses on the defensive side of VRSs and provides general insights that could be further exploited to broaden the frontier in the field. First, we introduce a suite of adversarial attacks against DNNs on top of VRSs, and defense strategies to counteract them. Next, we present an evaluation framework, named Visual Adversarial Recommender (VAR), to empirically investigate the performance of defended or undefended DNNs in various visually-aware item recommendation tasks. The results of large-scale experiments indicate alarming risks in protecting a VRS through the DNN robustification. Source code and data are available at https://github.com/sisinflab/Visual-Adversarial-Recommendation.

Model-based design of pharmaceutical manufacturing processes has received much interest in academia and industry. Model parameter uncertainties, however, might deteriorate the predicted process performance. Probability-based robust process design concepts as a countermeasure against uncertainties might be implemented. Here, parameter uncertainties are typically limited to Gaussian parameter distributions. However, parameter uncertainties derived with experimental data can be correlated and arbitrarily distributed. In our previous work, transformation techniques were combined with the point estimate method (PEM) to address non-Gaussian and correlated parameter distributions, but at the cost of additional nonlinearities and approximation errors. In this work, we take advantage of Gaussian mixture distributions (GMD) and decompose the parameter distribution into a finite set of Gaussian distributions using the Expectation-Maximization approach. Combining the GMD with the PEM ensures a proper and effective uncertainty quantification. The improved PEM algorithm is applied to a freeze-drying process (lyophilization) aiming for high-quality products with minimum processing time. Results obtained suggest that the novel GMD-PEM algorithm has the potential to outperform conventional robustification concepts regarding credibility and efficiency.

Robust generation of self-oscillation for mechanical systems in the presence of matched uncertainties/perturbations based on a two-relay controller, which naturally has a switched structure, is presented. Two scenarios are considered: the case when all the system states are available and the case when only output information is measured. In both scenarios, a continuous switched integral sliding modes controller is implemented to guarantee theoretically exact compensation of matched uncertainties/perturbations before the first switching. The chattering attenuation in the presence of fast parasitic dynamics can be realised by a switched gain strategy of the super-twisting algorithm. In the output-based scenario, the observer and the controller are initiated sequentially. A super twisting-based observer is used, guaranteeing theoretically exact reconstruction of the states before the controller is activated. The observer gains are also switched to diminish the chattering. Once the observer has converged, the reconstructed state is used in the controller to achieve the control objective. The observer and controller are designed to attain convergence before the first switching.

The goal of classification is to classify new objects into one of the several known populations. A common problem in most of the existing classifiers is that they are very much sensitive to outliers. To overcome this problem, several author’s attempt to robustify some classifiers including Gaussian Bayes classifiers based on robust estimation of mean vectors and covariance matrices. However, these type of robust classifiers work well when only training datasets are contaminated by outliers. They produce misleading results like the traditional classifiers when the test data vectors are contaminated by outliers as well. Most of them also show weak performance if we gradually increase the number of variables in the dataset by fixing the sample size. As the remedies of these problems, an attempt is made to propose a highly robust Gaussian Bayes classifiers by the minimum β-divergence method. The performance of the proposed method depends on the value of tuning parameter β, initialization of Gaussian parameters, detection of outlying test vectors, and detection of their variable-wise outlying components. We have discussed some techniques in this paper to improve the performance of the proposed method by tackling these issues. The proposed classifier reduces to the MLE-based Gaussian Bayes classifier when β → 0. The performance of the proposed method is investigated using both synthetic and real datasets. It is observed that the proposed method improves the performance over the traditional and other robust linear classifiers in presence of outliers. Otherwise, it keeps equal performance.

This paper proposes a new robustification strategy of a primary H∞ controller based on a systematic selection of adequate fractional weights. Its robust performance (RP) margin is well enhanced wit...

Contamination detection in acoustic signals is crucial for their applicability in industrial surroundings and for robustification of state estimation algorithms which use them. This paper presents an adaptive solution which estimates nominal signal statistics and detects contamination by measuring the deviation of the obtained signal distribution from the expected one without requiring a state space model that is usually used in existing algorithms. Detecting a change in nominal working condition, this algorithm adapts to it by repeating the training procedure. Statistical deviation detection is conducted using Quantile-Quantile plot (QQ-plot) based decision scheme which calculates metrics in the form of mean squared error of deviation of the recorded signal from the expected distribution. This algorithm has been tested and verified on real acoustic signals obtained in noisy industrial surroundings. The results are compared to a well known T 2 multivariate control chart procedure.

In this paper, two active fault-tolerant control (AFTC) structures are proposed for multiplicative and additive actuator faults respectively. Considering that the traditional generalized internal model control (GIMC) changes the dynamic input-output relationship of the original system, more exact fault information is employed to design the robustification controller of GIMC with the help of fault diagnosis observers. Further, a feedforward compensation unit and a feedback compensation unit are respectively designed for multiplicative and additive faults, striving to approximately recover the system performance to the normal one. Finally, an air handling unit (AHU) system is provided to verify the effectiveness and superiority of the proposed structures.

Hybrid electric vehicle (HEV) control strategies are often designed around specific driving conditions. However, when driving conditions differ from the designed conditions, HEV performance can suffer. This paper develops a novel HEV energy management strategy (EMS) that is robust to uncertain driving conditions by augmenting a stochastic dynamic programing (SDP) controller with minimax dynamic programing (MDP). This combination of MDP and SDP has not previously been studied in the literature. The stochastic element uses a Markov chain model to represent driver behavior and is used to optimize the control for expected future driver behavior. The minimax element instead optimizes against potential worst-case (maximal cost) future driver behavior. The resulting EMS can be directly implemented on a vehicle. This method is demonstrated on a series hybrid electric bus model. Robustness to uncertain driving conditions is tested by simulating on a variety of heavy-duty vehicle drive cycles that differ from the drive cycle on which the EMS was trained. A single tuning parameter is used to balance the stochastic and minimax elements of the EMS, and a parametric study shows the effects of this tuning parameter. It was found that using minimax control could increase the vehicle fuel economy on multiple uncertain driving conditions, with a tradeoff of decreased fuel economy when the driving conditions match the designed conditions. That is, it offers an exchange of performance on the nominal driving conditions for performance on uncertain driving conditions.

When designing a clinical trial, borrowing historical control information can provide a more efficient approach by reducing the necessary control arm sample size while still yielding increased power. Several Bayesian methods for incorporating historical information via a prior distribution have been proposed, for example, (modified) power prior, (robust) meta-analytic predictive prior. When utilizing historical control borrowing, the prior parameter(s) must be specified to determine the magnitude of borrowing before the current data are observed. Thus, a flexible prior is needed in case of heterogeneity between historic trials or prior data conflict with the current trial. To incorporate the ability to selectively borrow historic information, we propose a Bayesian semiparametric meta-analytic-predictive prior. Using a Dirichlet process mixture prior allows for relaxation of parametric assumptions, and lets the model adaptively learn the relationship between the historic and current control data. Additionally, we generalize a method for estimating the prior effective sample size (ESS) for the proposed prior. This gives an intuitive quantification of the amount of information borrowed from historical trials, and aids in tuning the prior to the specific task at hand. We illustrate the effectiveness of the proposed methodology by comparing performance between existing methods in an extensive simulation study and a phase II proof-of-concept trial in ankylosing spondylitis. In summary, our proposed robustification of the meta-analytic-predictive prior alleviates the need for prespecifying the amount of borrowing, providing a more flexible and robust method to integrate historical data from multiple study sources in the design and analysis of clinical trials.

The paper examines the sensitivity for the solution of linear programming problems using Bayesian techniques, when samples for the coefficients of the objective function are uncertain. When data is available, we estimate the solution of the linear program and provide statistical measures of uncertainty through the posterior distributions of the solution in the light of the data. When data is not available, these techniques examine the sensitivity of the solution to random variation in the coefficients of the linear problem. The new techniques are based on two posteriors emerging from the inequalities of Karush–Kuhn–Tucker conditions. The first posterior is asymptotic and does not require data. The second posterior is finite-sample-based and is used whenever data is available or if random samples can be drawn from the joint distribution of coefficients. A by-product of our framework is a robust solution. We illustrate the new techniques in two empirical applications to the case of uncertain Data Envelopment Analysis efficiency, involving two large samples, of US commercial banks and a sample of European commercial banks regulated by the Single Supervisory Mechanism. We analyse whether some pre-determined criteria, associated with size and new supervisory framework, can adequately affect the solution of linear program. The results provide evidence of substantial improvements in statistical structure with respect to sensitivities and robustification. Our methodology can serve as a consistency check of the statistical inference for the solution of linear programming problems in efficiency under uncertainty in data.

The robustification parameter, which balances bias and robustness, has played a critical role in the construction of sub-Gaussian estimators for heavy-tailed and/or skewed data. Although it can be tuned by cross-validation in traditional practice, in large scale statistical problems such as high dimensional covariance matrix estimation and large scale multiple testing, the number of robustification parameters scales with the dimensionality so that cross-validation can be computationally prohibitive. In this paper, we propose a new data-driven principle to choose the robustification parameter for Huber-type sub-Gaussian estimators in three fundamental problems: mean estimation, linear regression, and sparse regression in high dimensions. Our proposal is guided by the non-asymptotic deviation analysis, and is conceptually different from cross-validation which relies on the mean squared error to assess the fit. Extensive numerical experiments and real data analysis further illustrate the efficacy of the proposed methods

Frontiers in Bioengineering and Biotechnology 9, 685323 (2021). doi:10.3389/fbioe.2021.685323, Published by Frontiers Media, Lausanne

Construction crane vessels make use of dynamic positioning (DP) systems during the installation and removal of offshore structures to maintain the vessel's position. Studies have reported cases of instability of DP systems during offshore operation caused by uncertainties, such as mooring forces. DP “robustification” for heavy lift operations, i.e., handling such uncertainties systematically and with stability guarantees, is a long-standing challenge in DP design. A new DP method, composed by an observer and a controller, is proposed to address this challenge, with stability guarantees in the presence of uncertainties. We test the proposed method on an integrated cranevessel simulation environment, where the integration of several subsystems (winch dynamics, crane forces, thruster dynamics, fuel injection system etc.) allow a realistic validation under a wide set of uncertainties.

Autonomous vehicles increasingly rely on cameras to provide the input for perception and scene understanding and the ability of these models to classify their environment and objects, under adverse conditions and image noise is crucial. When the input is, either unintentionally or through targeted attacks, deteriorated, the reliability of autonomous vehicle is compromised. In order to mitigate such phenomena, we propose DriveGuard, a lightweight spatio-temporal autoencoder, as a solution to robustify the image segmentation process for autonomous vehicles. By first processing camera images with DriveGuard, we offer a more universal solution than having to re-train each perception model with noisy input. We explore the space of different autoencoder architectures and evaluate them on a diverse dataset created with real and synthetic images demonstrating that by exploiting spatio-temporal information combined with multi-component loss we significantly increase robustness against adverse image effects reaching within 5-6% of that of the original model on clean images., Comment: 2021 IEEE Winter Conference on Applications of Computer Vision Workshops (WACVW)

In recent years, the observed antibody sequence space has grown exponentially due to advances in high-throughput sequencing of immune receptors. The rise in sequences has not been mirrored by a rise in structures, as experimental structure determination techniques have remained low-throughput. Computational modeling, however, has the potential to close the sequence–structure gap. To achieve this goal, computational methods must be robust, fast, easy to use, and accurate. Here we report on the latest advances made in RosettaAntibody and Rosetta SnugDock—methods for antibody structure prediction and antibody–antigen docking. We simplified the user interface, expanded and automated the template database, generalized the kinematics of antibody–antigen docking (which enabled modeling of single-domain antibodies) and incorporated new loop modeling techniques. To evaluate the effects of our updates on modeling accuracy, we developed rigorous tests under a new scientific benchmarking framework within Rosetta. Benchmarking revealed that more structurally similar templates could be identified in the updated database and that SnugDock broadened its applicability without losing accuracy. However, there are further advances to be made, including increasing the accuracy and speed of CDR-H3 loop modeling, before computational approaches can accurately model any antibody.

Crafting adversarial inputs for attacks on neural networks and robustification against such attacks have continued to be a topic of keen interest in the machine learning community. Yet, the vast majority of work in current literature is only empirical in nature. We present a novel viewpoint on adversarial attacks on recurrent neural networks (RNNs) through the lens of dynamical systems theory. In particular, we show how control theory-based analysis tools can be leveraged to compute these adversarial input disturbances, and obtain bounds on how they impact the neural network performance. The disturbances are computed dynamically at each time-step by taking advantage of the recurrent architecture of RNNs, thus making them more efficient compared to prior work, as well as amenable to ‘real-time’ attacks. Finally, the theoretical results are supported by some illustrative examples.

This paper proposes a framework of signed small-gain criteria which are automatically amenable to asymmetry of gain-type dissipation with respect to equilibria. In theory and practice, the asymmetry has been avoided via over-bounding by symmetrization, which often results in unreasonably demanding or negative answers in stability, stabilization, robustness and robustification of dynamical systems. Natural systems are composed in accordance with the principle of conservation, and employ mechanisms which are more versatile than linearity. Exact energy balance is exhibited through nonlinear mechanisms, but it is not symmetric with respect to equilibria, due to the nonlinearity. Stability and robustness embedded in such energy balance cannot be detected after symmetrization. The signed small-gain methodology proposed in this paper provides powerful tools that absorb the asymmetry automatically in establishing integral input-to-stability of dynamical networks.

We study approaches to the robust model-based design of experiments in the context of maximum-likelihood estimation. These approaches provide robustification of model-based methodologies for the design of optimal experiments by accounting for the effect of the parametric uncertainty. We study the problem of robust optimal design of experiments in the framework of nonlinear least-squares parameter estimation using linearized confidence regions. We investigate several well-known robustification frameworks in this respect and propose a novel methodology based on multi-stage robust optimization. The proposed methodology aims at problems, where the experiments are designed sequentially with a possibility of re-estimation in-between the experiments. The multi-stage formalism aids in identifying experiments that are better conducted in the early phase of experimentation, where parameter knowledge is poor. We demonstrate the findings and effectiveness of the proposed methodology using four case studies of varying complexity.

The fifth-generation (5G) communication system is supposed to include a huge amount of mobile data traffic and number of wireless connections, for improved spectrum efficiency, quality of service (QoS), reliability, and security. This paper studies channel estimation techniques in multicarrier orthogonal frequency-division-multiplexing (OFDM) 5G wireless communication systems in Rayleigh and Rician channels. Proposed M-estimator based channel estimation technique in comparison with classical least squares (LS) and linear minimum mean-squared error (LMMSE) estimation is studied and analyzed. An M-estimator is proposed for robustification of channel estimation technique in Rayleigh and Rician fading channels. Simulation results are provided and the comparison shows the performance gains achieved by the proposed technique over LS and MMSE estimation in multicarrier OFDM 5G wireless communication systems.

International audience; A successful implementation of Model Predictive Control (MPC) requires appropriately tuned parameters. In This paper an Artificial Neural Network (ANN) based approach is presented. To build the data learning base of the ANN, we adopted the Particle Swarm Optimisation (PSO) method, and we used an Online Sequential Extreme-Learning-Machine (OS-ELM) algorithm to learn the ANN. The objective of this work is to to ensure robustness of the closed-loop system against neglecting dynamics, disturbances and sensor noises, and also show that good tuning of MPC parameters allows to reach closed-loop stability without using robustification approaches in addition to MPC. The effectiveness of our approach has been emphasized by comparing the obtained performances to Generalized Predictive Control (GPC) approach without robustification approach, and also to a robustified (GPC) using Youla parametrisation in the presence of disturbances.

International audience; A successful implementation of Model Predictive Control (MPC) requires appropriately tuned parameters. This paper presents a novel tuning approach based on Artificial Neural Network (ANN). To build the data learning base of the ANN, we adopted the Particle Swarm Optimisation (PSO) method, and we used the reliable algorithm, Online Sequential Extreme-Learning-Machine (OS-ELM) to learn the ANN. The objective of this work is to show that good tuning of MPC parameters makes it possible to reach closed-loop stability and ensure robustness against disturbances and sensor noises, without using robustification approaches in addition to MPC. The effectiveness of our approach is brought to light by comparing the obtained performances to other MPC tuning approaches without disturbances, and also to a robustified Generalized Predictive Control (GPC) using Youla parametrisation in the presence of disturbances.

© 2020 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. In this paper, a robust economic model predictive controller is proposed for a water transport network. Considering that forecast of demand is required in the drinking water network (DWN) for future predictions of states in the MPC formulation, robustification of the proposed controller is undertaken considering demand uncertainties as unknown but bounded in a zonotopic set. Based on this uncertainty description, a robust MPC controller is designed that ensures that robust constraint satisfaction, stability and performance under uncertainties is guaranteed. The proposed approach is satisfactorily tested in a part of the Barcelona water transport network.

This paper presents results into a new regulation scheme using a PID and immersion invariance controller for robust stabilization and tracking in rotational-translational actuator (RTAC) system. The stabilization and tracking control of the RTAC system is first demonstrated by application of the immersion invariance control law acting alone before the utilization of the designed control law for robustification of an existing PID controller for the RTAC system. In the nonrobust PID stabilization experiment, settling times of about 15 s was recorded while the robust stabilization experiments with the immersion invariance controller in the loop had 10 s settling time. For the nonrobust PID tracking experiment a noticeable overshoot was recorded of about 36.27% and settling times of about 18 s. Robust tracking experiments with the immersion invariance controller in the loop saw zero overshoot and 8 s settling time.

This work tries to contribute to the design of legged robots with capabilities boosted through thrusterassisted locomotion. Our long-term goal is the development of robots capable of negotiating unstructured environments, including land and air, by leveraging legs and thrusters collaboratively. These robots could be used in a broad number of applications including search and rescue operations, space exploration, automated package handling in residential spaces and digital agriculture, to name a few. In all of these examples, the unique capability of thruster-assisted mobility greatly broadens the locomotion designs possibilities for these systems. In an effort to demonstrate thrusters effectiveness in the robustification and efficiency of bipedal locomotion gaits, this work explores their effects on the gait limit cycles and proposes new design paradigms based on shaping these center manifolds with strong foliations. Unilateral contact force feasibility conditions are resolved in an optimal control scheme.

In this paper we deal with cutting plane approaches for robust optimization. Such approaches work iteratively by solving a robust problem with reduced uncertainty set (robustification step) and determining a worst-case scenario in each iteration (pessimization step) which is then added to the reduced uncertainty set. We propose to enhance this scheme by solving the robustification and/or the pessimization step not exactly, but only approximately, that is, until an improvement to the current solution is possible. The resulting iterative approach is called approximate cutting plane approach. We prove that convergence to an optimal solution for approximate cutting plane approaches is still guaranteed under similar assumptions as for classical cutting plane approaches, in which both robustification and pessimization problem are solved exactly in each iteration. Experimentally, we investigate robust mixed integer linear optimization problems for mixed-integer polyhedral uncertainty sets of different difficulties. Solving the robustification or pessimization problem only approximately increases the number of iterations. Nevertheless, our results show that the approximate cutting plane approach becomes more efficient, in particular, if the robustification or the pessimization problem is hard.

Widespread application of computer vision systems in real world tasks is currently hindered by their unexpected behavior on unseen examples. This occurs due to limitations of empirical testing on finite test sets and lack of systematic methods to identify the breaking points of a trained model. In this work we propose semantic adversarial editing, a method to synthesize plausible but difficult data points on which our target model breaks down. We achieve this with a differentiable object synthesizer which can change an object’s appearance while retaining its pose. Constrained adversarial optimization of object appearance through this synthesizer produces rare/difficult versions of an object which fool the target object detector. Experiments show that our approach effectively synthesizes difficult test data, dropping the performance of YoloV3 detector by more than 20 mAP points by changing the appearance of a single object and discovering failure modes of the model. The generated semantic adversarial data can also be used to robustify the detector through data augmentation, consistently improving its performance in both standard and out-of-dataset-distribution test sets, across three different datasets.

In this paper, a new approach to perfect control design is proposed. The novel application of polynomial degrees of freedom in calculation of generalized matrix inverse has enabled a new branch of perfect control-oriented robust scenarios. Crucially, the obtained closed-loop system has revealed to be internally time-varying even though the general input-output behavior is still time-invariant. Simulation instances made in Matlab/Simulink environment show some interesting peculiarities covering the perfect control speed and energy.

Metalearning is becoming an increasingly important methodology for extracting knowledge from a database of available training datasets to a new (independent) dataset. While the concept of metalearning is becoming popular in statistical learning and readily available also for the analysis of economic datasets, not much attention has been paid to its limitations and disadvantages. To the best of our knowledge, the current paper represents the first illustration of metalearning sensitivity to data contamination by noise or outliers. For this purpose, we use various linear regression estimators (including highly robust ones) over a set of 24 datasets with economic background and perform a metalearning study over them as well as over the same datasets after an artificial contamination. The results reveal the whole process to remain rather sensitive to data contamination and some of the standard classifiers turn out to yield unreliable results. Nevertheless, using a robust classification method does not bring a desirable improvement. Thus, we conclude that the task of robustification of the whole metalearning methodology is more complex and deserves a systematic future research.

This study approaches a robustification method for a multi-robot network connectivity. Instead of the vertex connectivity, which is commonly used as a robustness index, here we consider the size of the connected component remaining after one robot has been removed from the network, and we propose a distributed control law for improvement and preservation of the remaining connected component size. Some conditions of a modified graph Laplacian eigenvalue are analyzed for the improvement and the preservation, and then the control strategy is composed using the Laplacian eigenvalue as an indicator of the remaining connected component size. From simulations, we observed that a multi-robot system with our control method achieves a convincing state regarding the trade-off between a network robustness and a coverage task performance.

This paper presents a novel yet efficient defense framework for segmentation models against adversarial attacks in medical imaging. In contrary to the defense methods against adversarial attacks for classification models which widely are investigated, such defense methods for segmentation models has been less explored. Our proposed method can be used for any deep learning models without revising the target deep learning models, as well as can be independent of adversarial attacks. Our framework consists of a frequency domain converter, a detector, and a reformer. The frequency domain converter helps the detector detects adversarial examples by using a frame domain of an image. The reformer helps target models to predict more precisely. We have experiments to empirically show that our proposed method has a better performance compared to the existing defense method.

By the use of online optimization, the profitability of chemical processes can be enhanced while meeting process-related, environmental and ecological constraints. Two well-known techniques to achieve an economically optimal operation of chemical processes are repetitive stationary optimization (RTO) combined with advanced control and direct economics optimizing control on a receding horizon (also called one-layer approach or dynamic RTO). Direct economics optimizing control can react faster to disturbances and is better suited for the optimization of transients between different products. On the other hand it is computationally demanding and requires sufficiently accurate dynamic models. In this paper, we discuss economics optimizing control of a very complex process, a multi-product transesterification reaction that is realized in a reactive distillation process. The process model, the specification of the economics optimizing controller, the implementation of the dynamic optimization, and the robustification of the controller by a multi-stage approach are discussed using simulation studies.

In this paper we propose a new stochastic consensus algorithm based on the introduction of a nonlinear transformation aimed at robustification with respect to noise influence. The introduced nonlinear transformation is selected according to the methodology of stochastic approximation and robust statistics. The proposed algorithm represents a general nonlinear stochastic consensus seeking scheme, not yet treated in the literature. It is proved, under general conditions, that the algorithm converges almost surely to consensus. The choice of the nonlinearity ensuring better robustness properties compared to the linear algorithm, in the sense of better convergence rate and lower sensitivity of the asymptotic consensus value, is discussed. Illustrative simulation results, demonstrating the obtained advantages, are also provided.

Machine Learning models are vulnerable to adversarial examples, viz. these models misclassify data points that are slightly deviated from the correctly classified examples drawn from the same data distribution. These adversarial examples are a result of and reveal inherent blind spots in our training algorithms. In this article, we proposed to build a blind robustification system which can be applied on all input samples such that it targets to retrieve the original labels of adversarial examples without causing changes to the prediction of non-adversarial samples. We first concretely propose an energy functional designed to target adversarial perturbation. We then derive its gradient descent flow equation using the Euler-Lagrange method. Following this, we derive an efficient discretization scheme for the proposed partial differential equations. Finally, we perform a thorough analysis of the performance of the proposed system by testing it for a state-of-art network and a universal bench-marking dataset.