Machine learning for people detection in guidance functionality of enabling health applications by means of cascaded SVM classifiers

Electronic assistance devices for visually impaired people often require pedestrian detection functionality. Due to the fact that these devices have only low computational processing capabilities, it is necessary to use realtime algorithms that meet these requirements.We thus propose application of a cascaded support vector machine (SVM) classifier with linear and combined cascading of SVMs based on histogram of oriented gradients (HOGs) and Fisher score preselection. Proposed algorithms are evaluated on the basis of the INRIA database and compared to the state of the art procedure with HOG-features in combination with a standard SVM.     

On the approximate discrete KLT of fractional Brownian motion and applications

This paper establishes connection between discrete cosine transform (DCT) and the discrete-time fractional Brownian motion process (dfBm). It is proved that the eigenvectors of the auto-covariance matrix of a dfBm can be approximated by DCT basis vectors in the asymptotic sense. This shows that DCT basis acts as discrete Karhunen–Loève transform (DKLT) for these processes in the approximate sense. Analytic perturbation theory of linear operators is used to prove this result. This result will be of great practical significance in applications where one is looking for an appropriate basis to work with signals that can perhaps be modeled as belonging to fBm processes. The utility of the proposed work has been illustrated with two real-life data (a) on compressive sampling based reconstruction of financial time-series and (b) in denoising gravitational wave event GW150914 data obtained from a binary black hole merger.     

Control rules extraction and parameters optimization of energy management for bus series-parallel AMT hybrid powertrain

A rule-based energy management strategy, that the control rules are extracted from acknowledged optimal algorithms and its control parameters are optimized offline and corrected online, for a series-parallel hybrid powertrain with an automatic mechanical transmission (AMT) is proposed in this paper to achieve near optimal fuel economy and battery state-of-charge (SOC) balance. Firstly, the dynamic programming (DP) global optimization method is applied to extract driving-mode transition rules and gear shifting rules. Furthermore, an instantaneous equivalent fuel consumption minimizing optimization method (ECMS) is utilized to determinate the engine torque distribution rules during its parallel driving mode. Then selected control parameters of driving-mode switching rules and torque split distribution are optimized based on genetic algorithm (GA) for further fuel consumption improvement. And the adaptive correction of optimized control parameters based on online driving cycle recognition method is discussed also. The simulation results show that this real-time rule-based energy management control strategy associated with the series of optimization approaches comprehensively can achieve a relatively close fuel consumption results to global optimal results and sustain the battery SOC balance after the end of driving cycle without much cycle-depending care.     

Orthogonal-polynomials-based integral inequality and its applications to systems with additive time-varying delays

Recently, a polynomials-based integral inequality was proposed by extending the Moon’s inequality into a generic formulation. By imposing certain structures on the slack matrices of this integral inequality, this paper proposes an orthogonal-polynomials-based integral inequality which has lower computational burden than the polynomials-based integral inequality while maintaining the same conservatism. Further, this paper provides notes on relations among recent general integral inequalities constructed with arbitrary degree polynomials. In these notes, it is shown that the proposed integral inequality is superior to the Bessel–Legendre (B–L) inequality and the polynomials-based integral inequality in terms of the conservatism and computational burden, respectively. Moreover, the effectiveness of the proposed method is demonstrated by an illustrative example of stability analysis for systems with additive time-varying delays.     

The obstacle problem of integro-partial differential equations with applications to stochastic optimal control/stopping problem

This paper is devoted to existence and uniqueness of minimal mild super solutions to the obstacle problem governed by integro-partial differential equations. We first study the well-posedness and local Lipschitz regularity of L^p solutions (p?≥?2) to reflected forward-backward stochastic differential equations (FBSDEs) with jump and lower barrier. Then we show that the solutions to reflected FBSDEs provide a probabilistic representation for the mild super solution via a nonlinear Feynman–Kac formula. Finally, we apply the results to study stochastic optimal control/stopping problems.     

Parametric delta operator Riccati equation and applications to semi-global stabilization in the presence of actuator saturation

In this paper, a parametric delta operator Riccati equation is established for low gain feedbacks of linear delta operator systems. Some properties for the parametric delta operator Riccati equation are given based on a parameter-dependent cost function. An explicit solution is also given for the delta operator parametric Riccati equation. Semi-global stabilization is described for a linear delta operator system with actuator saturation via low gain state and output feedback control laws. A numerical example is given to illustrate the effectiveness and potential for the developed techniques.     

Super-twisting observer-based sliding mode control with fuzzy variable gains and its applications to fully-actuated hexarotors

In this paper, we consider the super-twisting observer-based sliding mode control algorithm with fuzzy variable gains (STOSMC) for the fully-actuated hexarotor. Our hexarotor has full actuation due to six titled propellers that allows to control position and orientation (attitude) simultaneously, and resolves the singularity problem of the rotational matrix by using the quaternion modeling framework. We show that the proposed STOSMC for the hexarotor guarantees finite-time convergence of the estimation error and asymptotic stability of the hexarotor. In simulations, we demonstrate the nonsingularity and fully-actuated control performance of the hexarotor by considering extreme position and attitude control scenarios. Moreover, the simulation results show that the hexarotor achieves the fast and precise tracking performance to the desired position and the desired attitude and the chattering phenomenon is reduced compared with the fixed-gains observer-based super-twisting sliding mode control due to the fuzzy mechanism.     

Stabilization of periodic orbits of discrete-time dynamical systems using the Prediction-Based Control: New control law and practical aspects

This paper tackles in the stabilization of periodic orbits of nonlinear discrete-time dynamical systems with chaotic sets. The problem is approximated locally to the stabilization of linear time-periodic systems and the theory of modern control is applied to the Prediction-Based Control, resulting in a new control law. This control law sets all the Floquet multipliers of the stabilized orbit to zero, resulting in fast convergence of trajectories in its vicinity. Another important characteristic of the control law is that no previous knowledge about the periodic orbit is required for stabilization. Using numerical simulations, this control law was analysed and compared to an optimal Delayed Feedback Control evidencing its advantages in theoretical and practical aspects.     

Subset selection for visualization of relevant image fractions for deep learning based semantic image segmentation

Semantic image segmentation is a challenging problem from image processing where deep convolutional neural networks (CNN) have been applied with great success in the recent years. It deals with pixel-wise classification of an input image, dividing it into regions of multiple object classes. However, CNNs are opaque models. Given a trained CNN, it is hard to tell which information encoded in the input image is important for the network to perform segmentation. Such information could be useful to judge whether a trained network learned to segment in a plausible way or how its performance can be improved.For a trained CNN, we formulate an optimization problem to extract relevant image fractions for semantic segmentation. We try to identify a subset of pixels that contain the relevant information for the segmentation of one selected object class. In experiments on the Cityscapes dataset, we show that this is an easy way to gain valuable insight into a CNN trained for semantic segmentation. Looking at the relevant image fractions, we can identify possible limits of a trained network and draw conclusions about possible improvements.     

Event-triggered backstepping control for attitude stabilization of spacecraft

To decrease the communication frequency between the controller and the actuator, this paper addresses the spacecraft attitude control problem by adopting the event-triggered strategy. First of all, a backstepping-based inverse optimal attitude control law is proposed, where both the virtual control law and the actual control law are respectively optimal with respect to certain cost functionals. Then, an event-triggered scheme is proposed to realize the obtained inverse optimal attitude control law. By designing the event triggering mechanism elaborately, it is guaranteed that the trivial solution of the closed-loop system is globally exponentially stable and there is no Zeno phenomenon in the closed-loop system. Further, the obtained event-triggered attitude control law is modified and extended to the more general case when the disturbance torque cannot be ignored. It is proved that all states of the closed-loop system are bounded, the attitude error can be made arbitrarily small ultimately by choosing appropriate design parameters and the Zeno phenomenon is excluded in the closed-loop system. In the proposed event-triggered attitude control approaches, the control signal transmitted from the controller to the actuator is only updated at the triggered time instant when the accumulated error exceeds the threshold defined elaborately. Simulation results show that by using the proposed event-triggered attitude control approach, the communication burden can be significantly reduced compared with the traditional spacecraft control schemes realized in the time-triggered way.     

Compressed sensing for real measurements of quaternion signals

The paper concerns compressed sensing methods in the quaternion algebra. We prove that it is possible to uniquely reconstruct – by ?₁ norm minimization – a sparse quaternion signal from a limited number of its real linear measurements, provided the measurement matrix satisfies so-called restricted isometry property with a sufficiently small constant. We also provide error estimates for the approximated reconstruction of a non-sparse quaternion signal from noisy and noiseless data.     

Iterative learning control for boundary tracking of uncertain nonlinear wave equations

This work presents a framework of iterative learning control (ILC) design for a class of nonlinear wave equations. The main contribution lies in that it is the first time to extend the idea of well-established ILC for lumped parameter systems to boundary tracking control of nonlinear hyperbolic distributed parameter systems (DPSs). By fully utilizing the system repetitiveness, the proposed control algorithm is capable of dealing with time-space-varying and even state-dependent uncertainties. The convergence and robustness of the proposed ILC scheme are analyzed rigorously via the contraction mapping methodology and differential/integral constraints without any system dynamics simplification or discretization. In the end, two examples are provided to show the efficacy of the proposed control scheme.     

New modeling approach of secondary control layer for autonomous single-phase microgrids

In a microgrid (MG) topology, the secondary control is introduced to compensate for the voltage amplitude and frequency deviations, mainly caused by the inherent characteristics of the droop control strategy. This paper proposes an accurate approach to derive small signal models of the frequency and amplitude voltage at the point of common coupling (PCC) of a single-phase MG by analyzing the dynamics of the second-order generalized integrator-based frequency-locked loop (SOGI-FLL). The frequency estimate model is then introduced in the frequency restoration control loop, while the derived model of the amplitude estimate is introduced for the voltage restoration loop. Based on the obtained models, the MG stability analysis and proposed controllers’ parameters tuning are carried out. Also, this study includes the modeling and design of the synchronization control loop that enables a seamless transition from island mode to grid-connected mode operation. Simulation and practical experiments of a hierarchical control scheme, including traditional droop control and the proposed secondary control for two single-phase parallel inverters, are implemented to confirm the effectiveness and the robustness of the proposal under different operating conditions. The obtained results validate the proposed modeling approach to provide the expected transient response and disturbance rejection in the MG.     

Formation control and collision avoidance for a class of multi-agent systems

In this paper, a control scheme is proposed for a group of elliptical agents to achieve a predefined formation. The agents are assumed to have the same dynamics, and communication among the agents are limited. The desired formation is realized based on the reference formation and the mapping decision. In the control design, searching algorithms for both cases of minimum distance and tangents are established for each agent and its neighbors. In order to avoid collision, an optimal path planning algorithm based on collision angles, and a self-center-based rotation algorithm are also proposed. Moreover, randomized method is used to provide the optimal mapping decision for the underlying system. Two examples and analyses are presented to demonstrate the effectiveness and potential of the new control scheme.     

Necessary and sufficient criteria for asynchronous stabilization of Markovian jump systems

This paper is concerned with asynchronous stabilization for a class of discrete-time Markovian jump systems. The mode of designed controller is considered to be not perfectly synchronous with the activated mode of the Markovian jump system. In order to achieve the asymptotic stability with asynchronous controller, a conditional probability is introduced to describe the asynchronism of system and controller modes, which is dependent on the active system mode. Besides, due to the difficulty in acquiring all the mode transition probabilities in practice, the transition probabilities of the Markovian jump system and the controllers are supposed to be partially unknown. A necessary and sufficient condition is developed to guarantee the stochastic stability of the resultant closed-loop system and further extended to asynchronous stabilization with partially known transition probabilities. Finally, the effectiveness and advantages of the proposed methods are demonstrated by two illustrative examples.     

Modification of Mikhailov stability criterion for fractional commensurate order systems

In this paper we present the modification of the Mikhailov stability criterion for linear fractional commensurate order systems. The modification consists in determining the appropriate measure for the total argument change depending on the highest fractional order ${\alpha }_n=n\alpha$ of the system and not only on the integer n as stated in the literature. The validity of the result is illustrated by means of several examples.     

A class of fast fixed-time synchronization control for the delayed neural network

This paper deals with the synchronization control of a class of delayed neural networks using a fast fixed-time control theory. By employing Lyapunov stability theory, a novel sufficient criterion is derived such that two neural networks can be synchronized within a fixed-time. Compared with some existing results, the proposed controller can render two neural networks faster synchronized. A numerical example is given to demonstrate the effectiveness of the criterion.     

Evaluation of fault diagnosability for networked control systems subject to missing measurements

The evaluation of fault diagnosability for networked control systems subject to missing measurements is addressed in this paper. In particular, the missing probability is assumed to be affected by norm-bounded uncertainties. By considering noise and missing measurements, the quantitative diagnosability problem is investigated via the principle of weight. To quantify the effect of uncertain missing probabilities, an interval criterion is presented and uncertainty ratio is defined. Furthermore, a novel method is proposed to alleviate the computation task of evaluating fault diagnosability. The effectiveness of the proposed method is verified by a numerical example.     

Improved similarity-based modeling for the classification of rotating-machine failures

Similarity-based modeling (SBM) is a technique whereby the normal operation of a system is modeled in order to detect faults by analyzing their similarity to the normal system states. First proposed around two decades ago, SBM has been successfully used for fault detection in varied systems. In spite of this success, there is not much study performed in the literature regarding its design, that encompasses both similarity metrics and model training. This work aims at contributing with an in-depth study of SBM for fault detection considering these two design aspects. This is done in the context of proposing a novel system to identify rotating-machinery faults based on SBM, that is employed either as a standalone classifier or to generate features for a random forest classifier. New approaches for training the model and new similarity metrics are investigated. Experimental results are shown for the recently developed Machinery Fault Database (MaFaulDa) that has an extensive set of sequences and fault types, and for the Case Western Reserve University (CWRU) bearing database. Results for both databases indicate that the proposed techniques increase the generalization power of the similarity model and of the associated classifier, achieving accuracies of 98.5% on MaFaulDa and 98.9% on CWRU database.