This article addresses the event-triggered privacy preserving leader-follower consensus control problem for linear Multi-Agent Cyber-Physical Systems (MACPS) subject to cyber-attacks. A novel framework is proposed that combines virtual dynamics, event-triggered communication, and cyber-attack estimation to achieve output consensus while preserving agent privacy and system resilience against cyber-attacks. By introducing event-triggered virtual nodes, we enhance privacy of the agents while also managing communication resources efficiently by reducing data transmissions. To mitigate the effects of cyber-attacks, auxiliary systems are employed to estimate cyber-attack signals, enabling the design of a resilient control law to achieve leader-follower consensus in a Uniformly Ultimately Bounded (UUB) sense. Finally, Simulation results demonstrate the effectiveness of the proposed method in achieving leader-follower consensus despite cyber-attacks and communication limitations.

In this paper, the problem of attack detection and isolation for a fleet of naval vessels as a multi-agent Cyber-Physical System (CPS) will be addressed. Towards this, in the first step, a time-varying distributed formation control for a fleet of naval vessels as a network of leader-follower multi-agent systems using adaptive observer is presented. After solving formation control for the fleet of naval vessel, the detection and isolation of cyber-attacks for the system based on the obtained protocol will be considered. We demonstrate that the adaptive observers employed by each agent can identify cyber-attacks on communication channels originating from neighboring vessels. Additionally, through the development of a bank of adaptive observers that utilize received information from neighboring vessels and subsequent analysis of available residuals, we can isolate the compromised communication channels. Several cyber-attack scenarios will be discussed to show the performance of the proposed methodology for a fleet of naval vessels.

This paper deals with the detection and identification of the type of cyber-attack in switched Cyber-Physical Systems (CPS) operating under the synchronous switching conditions. The suggested approach involves an auxiliary system with a particular structure on the plant side, along with three switched observers on the Command and Control (C& C) side. To accomplish detection and identification objectives, the output of the auxiliary system needs to be communicated as well as real measurements of the system. Using the communicated information, three sets of residuals are obtained based on output estimation errors of two Switched Unknown Input Observers (SUIO) and a Switched Luenberger Observer (SLO). In the suggested approach, no secure channel or system is required and the least possible amount of information needs to be secured, which corresponds to the delay between the mode of the plant and auxiliary systems. Various attack scenarios such as covert attack, False Data Injection (FDI) attack on the input channels, FDI attack on the measurement channels and zero-dynamics attacks can be efficiently detected and identified by the proposed methodology. Simulation results are provided to demonstrate the effectiveness of the proposed scheme.

This paper investigates the vulnerability of marine vessels to cyber-attacks on their rudder servo systems. We present a state-space representation of the rudder servo system under cyber-attacks and propose a detection and identification methodology by using an interacting multiple-model unscented Kalman filter (IMM-UKF). An active-resilient control scheme is then developed that utilizes the multiple-model structure and dynamic reconfiguration to counter certain cyber-attack modes. We develop and propose a multiple-model control recovery methodology by utilizing the eigenstructure assignment method. By utilizing the proposed resilient control scheme for the rudder servo system, the eigenvalues of the vessel under cyber-attacks remain the same as those of the attack-free system. Consequently, the vessel can maintain and track its course and trajectory in the presence of cyber-attacks on the rudder servo system. In the conducted numerical case study, the effectiveness of the proposed multiple model resilient control scheme against cyber-attacks on the rudder servo system is investigated and demonstrated.

A new hybrid active noise control (ANC, HANC) system is proposed in this paper that is equipped with an input-power-controlled online secondary-path modeling (OSPM) subsystem. An FIR linear prediction filter (LPF) is newly included that takes the FIR supporting filter (SF) error as its desired signal and separates the remaining target narrowband noise component from all the other broadband noise components. Placed right after the LPF is the OSPM subsystem. The SF and LPF output signals, namely the target broadband and narrowband components that remain in the residual error are not only used to update the feedforward and feedback ANC (FFANC, FBANC) subcontroller, respectively, but are also adopted to control the power of the OSPM-exclusive auxiliary white Gaussian noise (AWGN) to pursue a trade-off between the OSPM quality and the AWGN contribution to the residual error. The OSPM error is utilized to simultaneously update not only the OSPM subsystem but also the SF and the LPF. Due to inclusion of the LPF, the adverse coupling effects among the FFANC, the FBANC and the OSPM is reduced substantially, leaving a possibility for improving the HANC overall convergence and noise reduction performance (NRP). Furthermore, preliminary steady-state analysis of the LPF is also conducted to reveal its properties and effectiveness. Extensive simulations with both synthetical and real settings are provided and conducted to verify that the proposed HANC system is superior to existing solutions.

In this paper, we investigate the problem of event-based covert attack detection in cyber-physical systems. We first investigate event-based covert attacks and discuss how covert attacks can be launched in event-based systems. We then use the capabilities of the event-based communication to detect these types of cyber-attacks. The notion of the auxiliary systems is used in our detection mechanism along with considering several anomaly detectors which could better detect anomalies in comparison to a single anomaly detector. Furthermore, prevalent assumptions in the literature about capabilities of the system (e.g. having secure channels or considering dynamics to be unknown to the attacker) are relaxed and finally, our approach is verified by providing several numerical case studies.

Given the important role of Electromechanical Actuators (EMAs) in the aviation industry, this paper aims to develop Prognosis and Health Monitoring (PHM) solutions for EMAs. We begin by analyzing the general configuration and architecture of EMAs and demonstrate that load torque oscillation induces amplitude modulation in the stator current. We also propose a relationship between two faults, namely spiral bevel gear and flex spline wear. Next, we model an EMA, including a brushless DC motor, inverter, gearbox, mechanical load, and other units. As a prerequisite for fault prediction, we address the estimation of two states: stator current and motor speed. We use a Particle Filter-based (PF) methodology to estimate these states and perform predictions. The prediction scheme involves forming an auxiliary state corresponding to fault degradation, based on which the remaining useful life (RUL) of the system is computed. Finally, we present extensive simulation results of the proposed methodology corresponding to various scenarios.

Abstract In this study, a robust feedforward hybrid active noise control (ANC) system with online secondary‐path modelling (SPM) is proposed that is capable of not only effectively suppressing the broadband and narrowband noise components but also tracking the secondary path (SP) variations. An finite impulse response online SPM subsystem as well as an efficient decoupling filter are included in the proposed feedforward hybrid ANC (HANC) system. The decoupling filter is a parallel‐form bandpass filter bank that consists of multiple bandpass filters that are derived from the second‐order infinite impulse response notch filters. It takes the residual noise as its input and separates the broadband component from the narrowband component, with the former used as not only a desired signal for the SPM but also as an error signal for updating the broadband sub‐controller, whereas the latter adapted simultaneously to scale the auxiliary white Gaussian noise and to update the narrowband sub‐controller. Extensive simulations are conducted with both the synthetic and real SPs as well as the synthetic and real noise signals that are generated by a large‐scale factory cutting machine (strand‐cutter) to demonstrate the advantages and effectiveness of the proposed feedforward HANC system. Comparisons are also demonstrated with the original HANC system as well as its directly extended version with an online SPM subsystem.

Second-order IIR adaptive notch filters (ANF) andbanks (ANFB) have been extensively utilized in many applications ranging from biomedical engineering to control systems. In this paper, a new efficient ANFB is proposed that is composed of two filter banks, namely a prefilter bank with cascaded IIR notch filters and a parallel structure IIR ANFB. The second-order IIR ANFs or filter bank cells in the proposed ANFB are updated by utilizing a modified normalized gradient (ModNG) algorithm. The proposed ANFB requires the same number of cells as does a conventional parallel form ANFB. However, the filter weight initialization drawback that the conventional parallel form ANFB inherently suffers from is significantly alleviated in the proposed ANFB. Equipped with the ModNG algorithm, the proposed ANFB yields considerably improved frequency estimation performance as compared to the conventional parallel form ANFB, for a wide range of frequencies and signal-to-noise ratios (SNR). Furthermore, preliminary steady-state analysis is provided for both the proposed and conventional ANFBs, which significantly enhances our understanding of their capabilities and features. Extensive case study simulations are conducted to demonstrate the superiority of the proposed ANFB over the conventional parallel form ANFB. The proposed ANFB is also applied to real noise and vibration signals to reveal its promising analytical and practical capabilities.

Fault diagnosis and prognosis are some of the most crucial functionalities in complex and safety-critical engineering systems, and particularly fault diagnosis, has been a subject of intensive research in the past four decades. Such capabilities allow for detection and isolation of early developing faults as well as prediction of fault propagation, which can allow for preventive maintenance, or even serve as a countermeasure to the possibility of catastrophic incidence as a result of a failure. Following a short preliminary overview and definitions, this article provides a survey of recent research on fault prognosis. Additionally, we report on some of the significant application domains where prognosis techniques are employed. Finally, some potential directions for future research are outlined.

This paper is mainly concerned with developing security indices for linear cyber-physical systems (CPS). The approaches for computing security (and consequently vulnerability analysis) of CPS in the literature are based on algebraic methods and system matrices. In this paper, for the first time in the literature we formally address the security index analysis and computation from a geometric system theory perspective. This point of view enables one to develop an algorithm for computing an upper bound on the security index having a linear time complexity with respect to dimension of the system (i.e., O(n)). This is a significant improvement compared to the currently available approaches in the literature that have polynomial time complexity. Unlike the approaches in the literature our methodology does not need any restriction on the representation of the system. Moreover, the geometric approach provides a tool to formally analyze the attack signals injected to the CPS by introducing a new type of attack that is more sophisticated than zero dynamic attacks. Finally, we illustrate our proposed methodology through a numerical example.

Electro hydraulic servo valves (EHSV) constitute as core parts of many hydraulic actuators such as multi-functional spoilers (MFS) in aircraft systems. Their continuous health monitoring is important for the overall flight safety as well as economizing down times and repairs. Hence, diagnosis, prognosis and health monitoring (DPHM) of EHSVs is addressed in this paper for multiple modes of degradations that might be present in the system. The objective of this paper is to demonstrate how each mode of degradation may be isolated and estimated, and how the remaining useful life (RUL) of the EHSV can be assessed for a critically safe component under strict safety regulations. In this paper Bayesian tracking and particle filters are utilized to address three main diagnosis and prognosis problems, namely `isolation', `estimation', and the `remaining useful life'. Several case studies simulations have been provided to demonstrate and illustrate the capabilities of our proposed methodologies.

In this paper, a fault detection and isolation (FDI) methodology based on an immune system (IS) inspired mechanism known as the dendritic cell algorithm (DCA) is developed and implemented. Our proposed DCA-based FDI methodology is then applied to a well-known wind turbine test model. The proposed DCA-based scheme performs both detection as well as isolation of sensor faults given dual sensor redundancy, unlike other works in the literature that only address the fault detection problem and rely on analytical redundancy approach for accomplishing the fault isolation task. A nonparametric statistical comparison test is also performed to compare the performance of the DCA-based FDI scheme with another IS-based scheme known as the negative selection algorithm. Through extensive simulation case study scenarios the capabilities and performance of our proposed methodologies have been fully demonstrated and justified.

In this paper, a novel computationally intelligent-based electrocardiogram (ECG) signal classification methodology using a deep learning (DL) machine is developed. The focus is on patient screening and identifying patients with paroxysmal atrial fibrillation (PAF), which represents a life threatening cardiac arrhythmia. The proposed approach operates with a large volume of raw ECG time-series data as inputs to a deep convolutional neural networks (CNN). It autonomously learns representative and key features of the PAF to be used by a classification module. The features are therefore learned directly from the large time domain ECG signals by using a CNN with one fully connected layer. The learned features can effectively replace the traditional ad hoc and time-consuming user's hand-crafted features. Our experimental results verify and validate the effectiveness and capabilities of the learned features for PAF patient screening. The main advantages of our proposed approach are to simplify the feature extraction process corresponding to different cardiac arrhythmias and to remove the need for using a human expert to define appropriate and critical features working with a large time-series data set. The extensive simulations and case studies conducted indicate that combining the learned features with other classifiers will significantly improve the performance of the patient screening system as compared to an end-to-end CNN classifier. The effectiveness and capabilities of our proposed ECG DL classification machine is demonstrated and quantitative comparisons with several conventional machine learning classifiers are also provided.

High operational and maintenance costs represent as major economic constraints in the wind turbine (WT) industry. These concerns have made investigation into fault diagnosis of WT systems an extremely important and active area of research. In this paper, an immune system (IS) inspired methodology for performing fault detection and isolation (FDI) of a WT system is proposed and developed. The proposed scheme is based on a self nonself discrimination paradigm of a biological IS. Specifically, the negative selection mechanism [negative selection algorithm (NSA)] of the human body is utilized. In this paper, a hierarchical bank of NSAs are designed to detect and isolate both individual as well as simultaneously occurring faults common to the WTs. A smoothing moving window filter is then utilized to further improve the reliability and performance of the FDI scheme. Moreover, the performance of our proposed scheme is compared with another state-of-the-art data-driven technique, namely the support vector machines (SVMs) to demonstrate and illustrate the superiority and advantages of our proposed NSA-based FDI scheme. Finally, a nonparametric statistical comparison test is implemented to evaluate our proposed methodology with that of the SVM under various fault severities.

In the area of robust control, fault diagnosis, and fault tolerant control of linear systems, many fundamental problems can be recast as H∞, I1and generalized H2control frameworks leading to the so-called mixed norm or multiobjective optimization problems. This paper develops a new linear matrix inequality (LMI) approach to the problems of event-triggered multiobjective synthesis of feedback controllers and fault diagnosis filters through a unified framework. Toward this end, at first a general problem known as event-triggered integrated fault detection, isolation and control (E-IFDIC) is defined. By utilizing a filter to represent, characterize, and specify the E-IFDIC module, a multiobjective formulation of the problem is developed based on H∞, H-, I1and generalized H2performance criteria. It is shown that when an event-triggered strategy is applied to both the sensor and E-IFDIC module, the amount of data that is sent through the sensor-to-E-IFDIC module and E-IFDIC module-to-actuator channels are dramatically reduced. A set of 'MI feasibility conditions is derived to ensure the solvability of the problem, as well as to simultaneously obtain the E-IFDIC module parameters and the event-triggered conditions. Finally, it is shown that certain existing problems in the fields of time and event-triggered control and fault diagnosis can be considered as special cases of our proposed methodology. Two industrial case studies are also provided to illustrate and demonstrate the effectiveness of our proposed design methodology when compared with available work in the literature.

In this paper, a novel hybrid architecture is proposed for developing a prognosis and health monitoring methodology for nonlinear systems through integration of model-based and computationally intelligent-based techniques. In our proposed framework, the well-known particle filters (PFs) method is utilized to estimate the states as well as the health parameters of the system. Simultaneously, the system observations are predicted through an observation forecasting scheme that is developed based on neural networks (NNs) paradigms. The objective is to construct observation profiles that are to be used in future time horizons. Our proposed online training that is utilized for observation forecasting enables the NNs models to track nonergodic changes in the profiles that are present due to presence of hidden damage affecting the system health parameters. The forecasted observations are then utilized in the PFs to predict the evolution of the system states as well as the health parameters (which are considered to be time-varying due to effects of degradation and damage) into future time horizons. Our proposed hybrid architecture enables one to select health signatures for determining the remaining useful life of the system or its components not only based on the system observations but also by taking into account the system health parameters that are not physically measurable. Our proposed hybrid health monitoring methodology is constructed and developed by invoking a special framework where implementation of the observation forecasting scheme is not dependent on the structure of the utilized NNs model. In other words, changing the network structure will not significantly affect the prediction accuracy associated with the entire health prediction scheme. To verify and validate the above results and as a case study, our proposed hybrid approach is applied to predict the health condition of a gas turbine engine when it is affected by and subjected to fouling and erosion degradation and fault damages.

In this paper, we propose a new solar radiation (insolation) prediction method based on constructive neural networks (CoNN). The CoNN training process starts with a single hidden unit and then adds one hidden unit at a time when needed to enhance the network mapping capability to eventually obtain an effective and minimal neural network (NN). Historical insolation data, historical and forecasted meteorological variables (temperature, humidity, wind speed, and weather type), and the time parameter are used as input information to the CoNN. The proposed method is applied to real and actual dataset of hourly insolation data to demonstrate the forecasting superiority of our proposed approach over the naive persistence model (NPM), the autoregressive (AR) model, and the fixed-size feedforward NN (FFNN). Moreover, it has been revealed through extensive simulation studies that inclusion of a proper number of past weather variables and time parameter into the NN input vectors can further improve the prediction accuracy.

In this paper, a novel electrocardiogram (ECG) signal classification and patient screening method is developed. The focus is on identifying patients with paroxysmal atrial fibrillation (PAF) which is a life threatening cardiac arrhythmia. The proposed approach uses the raw ECG signal as the input and automatically learns the representative features for PAF to be used by a classification mechanism. The features are learned directly from the time domain ECG signals by using a Convolutional Neural Network (CNN) with one fully connected layer. The learned features can replace the hand-crafted features and our experimental results indicate the effectiveness of the learned features in patient screening. The experimental results indicate that combining the learned features with other classifiers will improve the performance of the patient screening system as compared to an End-to-End convolutional neural network classifier. The major characteristics of the proposed approach are to simplify the process of feature extraction for different cardiac arrhythmias and to remove the need for using a human expert to specify the appropriate features. The effectiveness of the proposed ECG classification method is demonstrated through performing extensive simulation studies.

In this paper, an online sensor fault detection and isolation (FDI) scheme is proposed based on an emerging Artificial Immune System (AIS) algorithm, namely Dendritic Cell Algorithm (DCA). Our proposed methodology is utilized in a distributed manner in order to perform sensor FDI in complex systems. The proposed methodology is then applied to a wind turbine benchmark model in order to demonstrate its capabilities.

In this paper, fault prognosis of aircraft jet engines are considered using computationally intelligent-based methodologies to ensure flight safety and performance. Two different dynamic neural networks namely, the nonlinear autoregressive neural networks with exogenous input (NARX) and the Elman neural networks are developed and designed for this purpose. The proposed dynamic neural networks are designed to capture the dynamics of two main degradations in the jet engine, namely the compressor fouling and the turbine erosion. The health status and condition of the engine is then predicted subject to occurrence of these deteriorations. In both proposed approaches, two scenarios are considered. For each scenario, several neural networks are trained and their performance in predicting multi-flights ahead turbine output temperature are evaluated. Finally, the most suitable neural network for prediction is selected by using the normalized Bayesian information criterion model selection. Simulation results presented demonstrate and illustrate the effective performance of our proposed neural network-based prediction and prognosis strategies.

In a fault detection system, generating residuals is the first step in detecting faults. However, residuals are not the only element of a dependable fault detection system. A fault detection system is reliable when an appropriate residual evaluation criterion is used along with a suitable residual generation technique. In this paper, a new method for an adaptive threshold generation is proposed to improve evaluation of the residuals with application to a trajectory following of an unmanned mobile robot. The proposed solution is useful when local linear models are utilized as observers for residual generation. For this purpose, locally linear model tree algorithm equipped with an external dynamics is applied as a powerful nonlinear identifier scheme to model the system. To demonstrate the capability of our proposed concept a complete model of a two wheeled mobile robot which is capable of implementing most possible faults in the system is developed. Detailed simulation results demonstrate the feasibility of our proposed methodology.

A novel methodology for detecting the transverse load distribution by fiber Bragg grating sensor (FBG) fabricated into the high birefringence fiber (Hi-Bi FBG sensor) is presented. The transverse force effects on the Hi-Bi FBG sensor area are modeled by evaluating the change of refractive indices along the fiber through the transfer matrix formulation method. Moreover, a genetic algorithm for reconstruction of the non-uniform applied anomaly from the reflected spectrum is developed. The presented method is verified through numerical simulations and the results show that our methodology can effectively represent and model the profile of the applied perturbation profile.

It is well known that for long-duration space missions, there is growing need for efficient utilization of telemetry data to enhance diagnostic performance and assist the less-experienced personnel in performing monitoring and diagnosis tasks. To address this need, we have, recently, developed a systematic and transparent fault diagnosis methodology within a hierarchical fault diagnosis framework for satellites formation flight. We developed our proposed hierarchical decomposition framework through a novel Bayesian network-based model, namely component dependence model (CDM). In this paper, we investigate the verification and validation of the CDM for fault diagnosis in satellites formation flight. We propose and develop a sensitivity analysis to verify the CDM by taking advantage of our systematic CDM development methodology. The proposed verification method satisfies the unique requirement of identifying CDM sensitivity when diagnostic performances of the algorithms that are deployed at one or more nodes of the CDM change. This implies that our verification approach and analysis are different from traditional sensitivity analysis that uses proportional scaling which is not applicable to the CDM methodology. Furthermore, in such analysis, a change in the model parameters under consideration is, typically, due to a change in the subjective judgment of an expert whose opinion is used in model development as opposed to the changes due to diagnostic performance variations. We demonstrate the proposed verification approach by using synthetic formation flight data, and show that our CDM development method does not lead to a fault diagnosis model that is sensitive to small variation in its parameters.

Current spacecraft health monitoring and fault-diagnosis practices involve around-the-clock limit-checking and trend analysis on large amount of telemetry data. They do not scale well for future multiplatform space missions due the size of the telemetry data and an increasing need to make the long-duration missions cost-effective by limiting the operations team personnel. The need for efficient utilization of telemetry data achieved by employing machine learning and reasoning algorithms has been pointed out in the literature for enhancing diagnostic performance and assisting the less-experienced personnel in performing monitoring and diagnosis tasks. In this paper, we develop a systematic and transparent fault-diagnosis methodology within a hierarchical fault-diagnosis framework for a satellites formation flight. We present our proposed hierarchical decomposition framework through a novel Bayesian network, whose structure is developed from the knowledge of component health-state dependencies. We have developed a methodology for specifying the network parameters that utilizes both node fault-diagnosis performance data and domain experts' beliefs. Our proposed model development procedure reduces the demand for expert's time in eliciting probabilities significantly. Our proposed approach provides the ground personnel with an ability to perform diagnostic reasoning across a number of subsystems and components coherently. Due to the unavailability of real formation flight data, we demonstrate the effectiveness of our proposed methodology by using synthetic data of a leader-follower formation flight architecture. Although our proposed approach is developed from the satellite fault-diagnosis perspective, it is generic and is targeted toward other types of cooperative fleet vehicle diagnosis problems.

In this paper, an optimal control design strategy for guaranteeing consensus achievement in a network of multiagent systems is developed. Minimization of a global cost function for the entire network guarantees a stable consensus with an optimal control effort. In solving the optimization problem, it is shown that the solution of the Riccati equation cannot guarantee consensus achievement. Therefore, a linearmatrix-inequality (LMI) formulation of the problem is used to address the optimization problem and to simultaneously resolve the consensus achievement constraint. Moreover, by invoking an LMI formulation, a semidecentralized controller structure that is based on the neighboring sets, i.e., the network underlying graph, can be imposed as an additional constraint. Consequently, the only information that each controller requires is the one that it receives from agents in its neighboring set. The global cost function formulation provides a deeper understanding and insight into the optimal system performance that would result from the global solution of the entire network of multiagent systems. Simulation results are presented to illustrate the capabilities and characteristics of our proposed multiagent team in achieving consensus.

In this paper, we investigate fault diagnosis and diagnosability in hybrid systems modeled by hybrid automata. Generally, in hybrid systems, there are discrete sensors generating discrete outputs available at the discrete-event system representation of the system and continuous sensors generating continuous outputs available at the continuous dynamics. We assume that there is a bank of residual generators (using continuous sensors) designed for the continuous dynamics of the system. We present a hybrid diagnosis approach in which faults are diagnosed by integrating the information generated by the residual generators and the information at the discrete-event system representation of the system. We investigate the diagnosability of faults in the hybrid diagnosis framework.

We have presented a framework in [6] for fault diagnosis of hybrid systems modeled by hybrid automata and investigated the diagnosability of failure modes in hybrid automata. In this framework, we assume that there is a bank of residual generators designed for the continuous dynamics of the system. We have developed a hybrid diagnosis approach in which faults are diagnosed by integrating the information generated by the residual generators and the information at the discrete-event system representation of the system. In this paper, we study the problem of residual generator selection in hybrid automata. Moreover, we investigate fault diagnosis of hybrid systems in the presence of unreliable residual generators generating false alarms or false silence signals.

This paper presents a robust fault detection and isolation (FDI) scheme for a general class of nonlinear systems using a neural-network-based observer strategy. Both actuator and sensor faults are considered. The nonlinear system considered is subject to both state and sensor uncertainties and disturbances. Two recurrent neural networks are employed to identify general unknown actuator and sensor faults, respectively. The neural network weights are updated according to a modified backpropagation scheme. Unlike many previous methods developed in the literature, our proposed FDI scheme does not rely on availability of full state measurements. The stability of the overall FDI scheme in presence of unknown sensor and actuator faults as well as plant and sensor noise and uncertainties is shown by using the Lyapunov's direct method. The stability analysis developed requires no restrictive assumptions on the system and/or the FDI algorithm. Magnetorquer-type actuators and magnetometer-type sensors that are commonly employed in the attitude control subsystem (ACS) of low-Earth orbit (LEO) satellites for attitude determination and control are considered in our case studies. The effectiveness and capabilities of our proposed fault diagnosis strategy are demonstrated and validated through extensive simulation studies.

The main objective of this paper is to develop a dynamic neural network-based fault detection and isolation (FDI) scheme for the Pulsed Plasma Thrusters (PPTs) of a satellite. The goal is to determine the occurrence of a fault in any one of the multiple thrusters that are employed in the attitude control subsystem of a satellite, and further to localize which PPT is faulty. In order to accomplish these objectives, a multilayer perceptron network embedded with dynamic neurons is proposed. Based on a given set of input-output data collected from the electrical circuit of the PPTs, the dynamic network parameters are adjusted to minimize the output estimation error. A Confusion Matrix approach is used to measure the effectiveness of our proposed dynamic neural network-based fault detection and isolation (FDI) scheme under various fault scenarios.

In this paper novel centralized and decentralized routing control strategies based on minimization of the worst-case queuing length are proposed. The centralized routing problem is formulated as an Hinfinoptimal control problem to achieve a robust routing performance in presence of multiple and unknown fast time-varying network delays. Unlike similar previous work in the literature the delays in the queuing model are assumed to be unknown and time-varying. A Linear Matrix Inequality (LMI) constraint is obtained to design a delay-dependent Hinfincontroller. The physical constraints that are present in the network are then expressed as LMI feasibility conditions. Our proposed centralized routing scheme is then reformulated in a decentralized frame work. This modification yields an algorithm that obtains the "fastest route", increases the robustness against multiple unknown time-varying delays, and enhances the scalability of the algorithm to large scale traffic networks. Simulation results are presented to illustrate and demonstrate the effectiveness and capabilities of our proposed novel dynamic routing strategies.

Noise signals generated by rotating machines such as diesel engines, cutting machines, fans, etc., may be modeled as noisy sinusoidal signals which can be successfully suppressed by narrowband active noise control (ANC) systems. In this paper, statistical performance of such a conventional filtered-x LMS (FXLMS)-based narrowband ANC system is investigated in detail. First, difference equations governing the dynamics of the system are derived in terms of convergence of the mean and mean squared estimation errors for the discrete Fourier coefficients (DFCs) of the secondary source. Steady-state expressions for DFC estimation mean square error (MSE) as well as the residual noise power are then developed in closed forms. A stability bound for the FXLMS in the mean sense is also derived. Extensive simulations of various scenarios are performed to demonstrate the validity of the analytical findings.

Rotating machines such as diesel engines, cutting machines, fans, etc. generate sinusoidal noise signals that may be successfully reduced by narrowband active noise control (ANC) systems. In this paper, performance analysis of such a typical filtered-X LMS (FXLMS) based narrowband ANC system equipped with an online secondary path modeling subsystem is conducted in detail. First, difference equations governing the dynamics of the FXLMS for secondary source synthesis and the LMS for secondary path estimation are derived in terms of convergence of both mean and mean square. Steady-state expressions for mean square error (MSE) as well as the remaining noise power are then developed in closed forms. Extensive simulations are performed to demonstrate the validity of the analytical results.

The goal of this paper is to develop a neural network-based scheme for fault detection and isolation in reaction wheels (actuators) of a satellite. To achieve this objective, three neural networks are developed for modeling the dynamics of a reaction wheel on all the three axes separately. A recurrent neural network with backpropagation training algorithm is considered for representing the highly nonlinear dynamics of the actuator. The capabilities and potential of the proposed neural network-based fault detection and isolation (FDI) methodology is investigated and a comparative study is conducted with the performance of a generalized Luenberger observer-based scheme. Simulation results demonstrate clearly the advantages of our proposed neural network scheme studied in this paper.

A new delay-dependent approach for robust control of multiple time-varying delayed systems with uncertain parameters is proposed. The internal stability of the proposed controller is shown by proposing a new Lyapunov-Krasovskii functional. The upper bound of the delay and its time-derivative are explicitly used in designing the controller. No constraint is imposed on the time-delay functions. Hence, the closed-loop system performance is more robust and less conservative in terms of tolerating the effects of time-varying delays. Moreover, the proposed controller does not rely on restrictive assumptions on the rate of change of time-delay function, (i.e. \tau\ < 1), which makes it applicable to fast time varying delayed systems. Finally, a robust state feedback controller is designed via Linear Matrix Inequality (LMI) technique. Unlike many previous methods, no parameter tuning is necessary to solve the resulting LMI conditions. Simulation results confirm that our proposed controller yields results that are more robust and less conservative as compared to existing methods in the literature.

This paper presents a method for detecting faults in a reaction wheel of an attitude control system of a spacecraft. A nonlinear observer is proposed as a residual generator to detect the anomalies and faults. To justify the need for the proposed complex nonlinear observer, a linear residual generation scheme is also considered as a benchmark and also for comparison. The linear observer is designed through linearization of the highly nonlinear dynamics of the reaction wheel. The performance and limitations of the linear observer subject to the three fault scenarios commonly present in the reaction wheel are investigated. The advantages and capabilities of applying the proposed nonlinear observer are demonstrated and emphasized through numerical simulations.

Formation flying is an emerging area in the Earth and space science and technology domain that utilize multiple inexpensive spacecraft by distributing the functionalities of a single platform among the miniature inexpensive platforms. Traditional spacecraft fault diagnosis and health monitoring practices that involve around-the-clock monitoring, threshold checking, and trend analysis of a large amount of telemetry data by human experts do not scale up well for multiple space platforms. In this paper a hierarchical fault detection and isolation (FDI) framework for spacecraft formation is proposed. Furthermore, fuzzy reasoning-based fault diagnosis for formation-level fault isolation related to attitude control is investigated. The proposed method has potential for acting as a mission enhancer by automating the fault diagnosis process for satellite formation flying missions.

In time-critical systems such as spacecraft systems, fault detection and isolation requirements are of paramount importance and necessity. This paper uses a second order nonlinear sliding mode observer to detect actuator faults in the attitude control subsystem of a satellite with four reaction wheels in a tetrahedron configuration. Through a postprocessing of residual signals it is shown how to isolate and reconstruct the faults in all four reaction wheels. Simulation results show that the proposed strategy can successfully detect, isolate and reconstruct reaction wheel faults.

The objective of this research is to develop to the proof-of-concept stage, a fault tolerant diagnosis system for the RADARSAT-1 attitude control system (ACS) telemetry. The proposed system is using computational intelligence (CI) to detect and isolate faults and also to infer cause of failures from the telemetry data time series history using functional models of satellite ACS. The proposed work is based on a distributed nonlinear, self-learning and self-adapting models (that can learn and improve themselves overtime) adjusting to the environment and constraints to which the real data is subjected. The key research and development issue is to create prototype models that will be able to integrate telemetry data and address the fault diagnosis problem without human intervention and expertise. The proposed work aims to support space industries' future interests in on-board fault diagnosis for next generation spacecraft by utilizing CI techniques as well as to help ground system operators in performing calibrations, anomaly detection, isolation and recovery, or testing of components.

This paper investigates development of a nonlinear Fault Detection and Isolation (FDI) strategy for redundant reaction wheels in the attitude control subsystem (ACS) of a satellite. Due to the coupling effects and dependencies in reaction wheels, the necessary condition for applying a standard geometric FDI approach is not satisfied. To remedy this problem, a set of detection filters are designed whereby through a combination of the residuals the FDI decision making is accomplished successfully. The simulation results presented demonstrate the effectiveness of our proposed FDI method.

In this paper, we develop a hybrid approach for fault diagnosis in hybrid systems assuming that multiple faults may occur in both continuous dynamics and at the discrete level. In this approach, a diagnoser is designed for a discrete-event-system (DES) abstraction of the system. The DES abstraction is modified such that the actual condition of the system always remains a subset of the condition estimate provided by the DES diagnoser. Whenever the DES diagnoser fails to provide a certain condition for the system (or it is deemed to be necessary), a continuous diagnoser consisting of a bank of residual generators becomes active. Since the continuous dynamics is not used at all times for diagnosis, our approach reduces the online computing requirements of the diagnosis system. We also model a typical spacecraft propulsion system as a hybrid automaton and explain how our results can be applied to it.

The objective of this work is to design a controller for a team of agents to accomplish a cohesive motion with consensus on the agreed upon output. A modified leader-follower structure for the team is considered. The desired output (command) is available to only the leader and followers are communicating with the leader and among themselves in a predefined topology. A semi-decentralized optimal control strategy is designed based on minimization of individual cost functions over a finite horizon using local information. Decentralization is achieved through incorporating interaction terms in the team members model. Minimization of the proposed cost function results in a modified consensus algorithm for a leader-follower structure when the leader command is available partially. Introduction of a corrective feedback from followers to the leader has potential advantage of improving robustness of the team to individuals faults. Finally, simulation results are presented to show effectiveness of our proposed method in achieving predefined requirements.

This paper presents a novel hybrid fault diagnosis approach to detect and estimate component faults in general nonlinear systems with full-state measurement. Unlike most existing fault diagnosis techniques, the proposed solution provides an integrated framework to simultaneously detect, isolate, and estimate the severity of faults in system components. The proposed solution consists of a bank of adaptive Neural Parameter Estimators (NPE) where each NPE in the bank is designed based on a separate parameterized fault model. Each NPE in the bank estimates its corresponding unknown Fault Parameter (FP) that is further used for fault detection and estimation purposes. Fast convergence and simple isolation policy are among the characteristic features of our proposed solution. Static neural network architecture and simple weight adaptation laws also make the proposed technique appropriate for real-time implementations. Simulation results reveal the effectiveness of the developed scheme in detecting, isolating and estimating faults in components of reaction wheel actuators of a 3-axis stabilized satellite even in presence of satellite disturbances.

This paper presents a robust Fault Detection and Isolation (FDI) scheme for a general nonlinear system using a neural network based observer. The nonlinear system is subject to state and sensor uncertainties and disturbances. A recurrent nonlinear-in-parameters neural network (NLPNN) is employed to identify the general unknown fault. The neural network weights are updated based on a modified backpropagation scheme. Unlike many previous methods, the proposed fault detection and isolation scheme does not rely on the availability of all state measurements. The stability of the overall fault detection approach in the presence of unknown faults as well as plant and sensor uncertainties is shown using Lyapunov's direct method. Stability analysis presented here imposes no restrictive assumptions on the system and/or the FDI algorithm. Magnetorquer type actuators that are commonly utilized in the satellite attitude control system is considered as a case study. The effectiveness of the proposed fault diagnosis strategy is demonstrated via simulations.

This paper presents an actuator Fault Detection and Isolation (FDI) scheme for nonlinear systems. A state space approach is used and a nonlinear-in-parameters neural network (NLPNN) is employed to identify the additive unknown fault. The FDI scheme is based on a hybrid model (composed of an analytical nominal model and a neural network model) of the nonlinear system. The nominal performance of the system in fault free operation is governed by analytical model whereas the uncertainties and unmodeled dynamics are accounted for by intelligent (neural network based) model. The neural network weights are updated based on a modified backpropagation scheme. The stability of the overall fault detection scheme is shown using Lyapunov's direct method. To evaluate the performance of the proposed fault detection scheme, FDI in a spacecraft attitude control systems with reaction wheel type of actuators is considered as a case study. Simulation results are presented to show the effectiveness of the proposed fault detection scheme.

The main objective of this paper is to develop a neural network-based fault detection and isolation scheme (FDI) for the attitude control subsystem (ACS) of a satellite. Towards this end, two neural network architectures are considered. First, a dynamic neural network residual generator is constructed based on the dynamic multilayer perceptron (DMLP) network to perform the detection task. A generalized embedded structure for the dynamic neuron model is considered in the DMLP network. Second, a static neural classifier is developed based on learning vector quantization (LVQ) network to be utilized for the isolation task. Based on a given set of input-output data collected from a 3-axis ACS of a satellite, the network parameters are adjusted to minimize a performance index specified by the output estimation error. The proposed neural FDI structure is applied to detect and isolate various faults in a high-fidelity nonlinear model of a satellite reaction wheel (RW), which is often used as an actuator in the ACS. The performance and capabilities of the proposed techniques are investigated and compared to a model-based observer residual generator that is used to detect various fault scenarios.

The objective of this paper is to develop a dynamic neural network scheme for fault detection and isolation (FDI) in the reaction wheels of a satellite. Specifically, the goal is to decide whether a bus voltage fault, a current loss fault or a temperature fault has occurred in one of the three reaction wheels and further to localize which wheel is faulty. In order to achieve these objectives, three dynamic neural networks are introduced to model the dynamics of the wheels on all three axes independently. Due to the dynamic property of the wheel, the architecture utilized is the Elman recurrent network with backpropagation learning algorithm. The effectiveness of this neural network-based FDI scheme is investigated and a comparative study is conducted with the performance of a generalized observer-based scheme. The simulation results have demonstrated the advantages of the neural network-based method proposed.

Human facial expression recognition (FER) has attracted much attention in recent years because of its importance in realizing highly intelligent human-machine interfaces. In this paper, we propose a new FER technique that utilizes the 2-D DCT of full size facial images and a decision tree with feedforward neural network (NN) based nodes. The first NN-based node of the decision tree is designated to separate one group of facial expressions with members "smile" and "surprise" from another group that contains "anger" and "sadness". This node can reduce the confusion between the category members of the two groups. Two NN-based nodes that follow the first node are established for each group to separate their two members. As a result, the original recognition problem with four categories is divided into three subproblems, each having only two members to distinguish. This work is the first trial to use NN, decision tree and 2-D DCT simultaneously within a single recognition task. To demonstrate the capability of the proposed recognition technique, we use two databases, including a recently constructed one, which contain 2-D front face images of 60 men and 60 women, respectively. Experimental results reveal that the new technique outperforms, on the whole, the simple vector matching and K-means based vector matching techniques and two recently developed methods using fixed-size and constructive neural networks. The mean recognition rates of the new technique have been found as high as 97.5% and 93.

Narrowband active noise control (ANC) systems have many real-life applications where the noise signals generated by rotating machines are modeled as sinusoidal signals in additive noise. However, when the timing signal sensor, such as a tachometer used to identify the signal frequencies, and the cosine wave generator contain errors, the frequencies of reference sinusoids fed to each ANC channel will then be different from the real primary noise signal frequencies. This difference is referred to as frequency mismatch (FM). In this paper, through extensive simulations, we first demonstrate that the performance capabilities of a conventional parallel form narrowband ANC system using the filtered-x LMS (FxLMS) algorithm degrades significantly even for an FM as small as 1 %. Convergence of the algorithm in the mean sense is also analytically investigated for a better understanding of its performance degradation. Next, we propose a new narrowband ANC system that successfully compensates for the performance degradations due to the FM. The amplitude/phase adjustment of reference sinusoids and the FM mitigations in the proposed system are performed simultaneously in a harmonic fashion such that the influence of the FM can be removed almost completely. Simulations as well as application to a real noise signal are provided to demonstrate the effectiveness of the proposed new system

The main objective of this paper is to develop a neural network-based residual generator for fault detection (FD) in the attitude control subsystem (ACS) of a satellite. Towards this end, a dynamic multilayer perceptron (DMLP) network with dynamic neurons is considered. The neuron model consists of a second order linear IIR filter and a nonlinear activation function with adjustable parameters. Based on a given set of input-output data pairs collected from the attitude control subsystem, the network parameters are adjusted to minimize a performance index specified by the output estimation error. The proposed dynamic neural network structure is applied for detecting faults in a reaction wheel (RW) that is often used as an actuator in the ACS of a satellite. The performance and capabilities of the proposed dynamic neural network is investigated and compared to a model-based observer residual generator design that is to detect various fault scenarios.

This paper presents an approach to simultaneous fault detection and isolation in the reaction wheel actuator of the satellite attitude control system. A model-based adaptive nonlinear parameter estimation technique is used based on a highly accurate reaction wheel dynamical model while each parameter is an indication of a specific type of fault in the system. The estimation is based on the nonlinear finite-memory filtering strategy that is solved for optimal estimation functions. To make the optimization feasible for on-line application, the optimal estimation functions are approximated by MLP neural networks thus reducing the functional optimization problem to a nonlinear programming problem, namely, the optimization of the neural weights. The well-known standard back-propagation algorithm and backpropagation through-time algorithm were employed inside the neural adaptation algorithms to obtain the required gradients. Simulation results show the effectiveness of the methodology for the proposed application.

In this paper, a constructive one-hidden-layer network is introduced where each hidden unit employs a polynomial function for its activation function that is different from other units. Specifically, both a structure level as well as a function level adaptation methodologies are utilized in constructing the network. The functional level adaptation scheme ensures that the "growing" or constructive network has different activation functions for each neuron such that the network may be able to capture the underlying input-output map more effectively. The activation functions considered consist of orthonormal Hermite polynomials. It is shown through extensive simulations that the proposed network yields improved performance when compared to networks having identical sigmoidal activation functions.

Facial expression recognition plays a vital role in realizing a highly intelligent human-machine interface, and has recently attracted much attention. In this paper, we propose a new facial expression recognition method that utilizes the 2D DCT, k-means algorithm and vector matching. This technique is based on two main intuitive ideas: (i) complicated facial expression categories such as "anger" and "sadness", may be divided into several subcategories with different subfeature spaces where the recognition task can be performed with higher accuracy and (ii) the k-means algorithm may be used to cluster these subcategories. A new image database with five facial expressions (neutral, smile, anger, sadness, surprise) of 60 women was constructed using a computationally efficient projection-based technique. Experimental results using the new database and an existing one (60 men) reveal that the new technique outperforms the standard vector matching technique and two recently developed methods using fixed-size and constructive one-hidden-layer neural networks. The mean recognition rate can be as high as 95% for the two databases.

A new technique for facial expression recognition is proposed, which uses the two-dimensional (2D) discrete cosine transform (DCT) over the entire face image as a feature detector and a constructive one-hidden-layer feedforward neural network as a facial expression classifier. An input-side pruning technique, proposed previously by the authors, is also incorporated into the constructive learning process to reduce the network size without sacrificing the performance of the resulting network. The proposed technique is applied to a database consisting of images of 60 men, each having five facial expression images (neutral, smile, anger, sadness, and surprise). Images of 40 men are used for network training, and the remaining images of 20 men are used for generalization and testing. Confusion matrices calculated in both network training and generalization for four facial expressions (smile, anger, sadness, and surprise) are used to evaluate the performance of the trained network. It is demonstrated that the best recognition rates are 100% and 93.75% (without rejection), for the training and generalizing images, respectively. Furthermore, the input-side weights of the constructed network are reduced by approximately 30% using our pruning method. In comparison with the fixed structure back propagation-based recognition methods in the literature, the proposed technique constructs one-hidden-layer feedforward neural network with fewer number of hidden units and weights, while simultaneously provide improved generalization and recognition performance capabilities.

This paper presents results on appearance-based three-dimensional (3-D) object recognition (3DOR) accomplished by utilizing a neural-network architecture developed based on independent component analysis (ICA). ICA has already been applied for face recognition in the literature with encouraging results. In this paper, we are exploring the possibility of utilizing the redundant information in the visual data to enhance the view based object recognition. The underlying premise here is that since ICA uses high-order statistics, it should in principle outperform principle component analysis (PCA), which does not utilize statistics higher than two, in the recognition task. Two databases of images captured by a CCD camera are used. It is demonstrated that ICA did perform better than PCA in one of the databases, but interestingly its performance was no better than PCA in the case of the second database. Thus, suggesting that the use of ICA may not necessarily always give better results than PCA, and that the application of ICA is highly data dependent. Various factors affecting the differences in the recognition performance using both methods are also discussed.

The objective of the paper is the application of an adaptive constructive one-hidden-layer feedforward neural networks (OHL-FNNs) to image compression. Comparisons with fixed structure neural networks are performed to demonstrate and illustrate the training and the generalization capabilities of the proposed adaptive constructive networks. The influence of quantization effects as well as comparison with the baseline JPEG scheme are also investigated. It has been demonstrated through several experiments that very promising results are obtained as compared to presently available techniques in the literature.

This paper presents two new modifications to the input-side training in constructive one-hidden-layer feedforward neural networks (FNNs). One is based on scaling of the network output error to which output of a hidden unit is expected to maximally correlate. Results from extensive simulations of many regression problems are then summarized to demonstrate that constructive FNNs generalization capabilities may be significantly improved by the new technique. The second contribution is a proposal for a new criterion for input-side weight pruning. This pruning technique removes redundant input-side weights simultaneously with the network constructive scheme, leading to a smaller network with comparable generalization capabilities. Simulation results are provided to illustrate the effectiveness of the proposed pruning technique.

Motivated by the well-known inverse dynamics control structure developed in the literature for flexible link manipulators, in this paper two multi-layer neural networks (NNs) are proposed to learn the nonlinearities of the system for achieving tip position trajectory tracking control for a single-link flexible manipulator. The re-defined output approach is used by feeding back this output to guarantee the minimum phase behavior of the resulting closed loop system. No a priori knowledge about the nonlinearities of the system is needed where the payload mass is also assumed to be unknown. The weights of the networks are adjusted using a modified online error backpropagation algorithm that is based on the propagation of output error, derivative of error and the tip deflection of the manipulator. The real-time controller is implemented on an experimental setup. The results achieved by the proposed neural network (NN) controller are compared experimentally with conventional PD and inverse dynamics controls to substantiate the advantages of our scheme and its promising potential.

The problem of controlling a discrete-time nonlinear non-minimum phase system is considered. An output re-definition strategy is developed which is applicable to a class of open-loop stable nonlinear systems whose input-output maps contain nonlinear terms from the output and linear terms from the input. No a priori knowledge about the nonlinearities of the system is required. The output re-definition scheme is based on first identifying the nonlinearities of the system using neural networks and then modifying the system zero dynamics. A stable/anti-stable factorization is performed on the zero dynamics of the system. The new output is re-defined using, the neural identifier and the stable part of the zero dynamics. A controller is then designed based on the new output whose zero dynamics are stable and can be inverted. Simulation results are resented to show the effectiveness of the proposed control scheme as compared to both linear and nonlinear conventional controllers.

The aim of this paper is to develop and implement a nonlinear adaptive control scheme for a single-link flexible manipulator. The controller is designed based on a discrete-time nonlinear model of the arm. The model is derived by using the forward difference method (Euler approximation). The output redefinition concept is then used so that the associated zero dynamics corresponding to the new output is guaranteed to be exponentially stable. An indirect adaptive linearizing controller is developed for the resulting minimum phase system where the "payload mass" is assumed to be unknown but its upper bound is assumed to be known a priori. The performance of the adaptively controlled closed-loop system is investigated by both numerical simulations and experimental results. The proposed controller is also compared experimentally with those of nonadaptive feedback linearization and conventional proportional derivative (PD) control strategies.

This paper deals with identification of a two-link flexible manipulator belonging to a class of multi-input, multi-output (MIMO) nonlinear systems, by using adaptive time delay neural networks (ATDNNs). Two neuro-dynamic identifiers are proposed. The capabilities of the proposed structures for representing the nonlinear input-output map of the flexible manipulator are shown analytically. Selection criteria for specifying the fixed structural parameters as well as the adaptation laws for updating the adjustable parameters of the networks are provided. During identification, the two-link flexible manipulator is under nonlinear control and the input-output data sets are generated for different desired trajectories. Simulation results reveal that the proposed neuro-dynamic structures are capable of successfully identifying a highly nonlinear system without any a priori information about the nonlinearities of the system and without any off-line training.

The problem of controlling the tip position of a flexible-link manipulator is considered. The control strategy is based on an output re-definition approach which is applicable to a class of non-minimum phase nonlinear systems whose nonlinearities appear in output terms in their input-ouput mappings and are open-loop stable. The controller is composed of a stabilizing joint PD controller and an output re-definition tracking controller. The output re-definition scheme is based on modifying the zero dynamics of the system using a stable/anti-stable factorization. The controller is then designed based on the new output whose zero dynamics are stable and can be inverted. Experimental results are also presented to show the effectiveness of the proposed control scheme as compared to a PD controller.

The problem of controlling a nonminimum phase nonlinear system with application to tip position control of a flexible-link manipulator is considered. An output re-definition strategy is developed which is applicable to a class of open-loop stable nonlinear systems whose input-output maps contain nonlinear terms from output and linear terms from input. No a priori knowledge about the nonlinearities of the system is required. The output re-definition scheme is based on first identifying the nonlinearities of the system using neural networks and then modifying the system zero dynamics. A stable/anti-stable factorization is performed on the zero dynamics of the system. The new output is re-defined using the neural identifier and the stable part of the zero dynamics. A controller is then designed based on the new output whose zero dynamics are stable and can be inverted. For the flexible-link manipulator case, the controller is composed of a stabilizing joint PD controller and an output re-definition tracking controller. Experimental and simulation results are presented to show the effectiveness of the proposed control scheme as compared to both linear and nonlinear conventional controllers.

Experimental evaluation of the performance of neural network-based controllers for tip position tracking of flexible-link manipulators is presented. A modified output re-definition approach is utilized to overcome the problem caused by the non-minimum phase characteristic of the flexible-link system. This modification is based on using minimum a priori knowledge about the system dynamics. The modified output redefinition approach requires a priori knowledge about the linear model of the system and no a priori knowledge about the payload mass. Four different neural network schemes are proposed. The neural networks are trained and employed as online controllers. The four proposed neural network controllers are implemented on a single flexible-link experimental test-bed. Experimental and simulation results are presented to illustrate the advantages and improved performance of the proposed tip position tracking controllers over conventional PD-type controllers in the presence of unmodeled dynamics.

This paper focuses on the experimental implementation of an observer-based decoupling control strategy for tip-position tracking of a class of multilink flexible manipulators. The control strategy is based on output redefinition and input-output decoupling that was studied by the authors in an earlier paper. Since the rates of change of flexible modes are required, a nonlinear observer is implemented to estimate these variables. Experimental results are given for the case of a two-link flexible manipulator that further confirm the theoretical and simulation results. The closed-loop performance under different observation schemes is evaluated and compared to the case when conventional methods are used.

This paper presents neural network based adaptive controllers for a single flexible-link manipulator. The control is designed by using the output re-definition approach. Three different neural network schemes are examined. None of the schemes requires full state measurements. The neural network controllers are implemented on a single flexible-link experimental test-bed. Experimental and simulation results are presented to illustrate the advantages and improved performance of the proposed tip position tracking controllers over the conventional PD-type controller.

In this paper, a nonlinear control strategy for tip position trajectory tracking of a class of structurally flexible multilink manipulators is developed. Using the concept of integral manifolds and singular perturbation theory, the full-order flexible system is decomposed into corrected slow and fast subsystems. The tip-position vector is similarly partitioned into corrected slow and fast outputs. To ensure an asymptotic tracking capability, the corrected slow subsystem is augmented by a dynamical controller in such a way that the resulting closed-loop zero dynamics are linear and asymptotically stable. The tracking problem is then redefined as tracking the slow output and stabilizing the corrected fast subsystem by using dynamic output feedback. Consequently, it is possible to show that the tip position tracking errors converge to a residual set of O(/spl epsiv//sup 2/), where /spl epsiv/ is the singular perturbation parameter. A major advantage of the proposed strategy is that the only measurements required are the tip positions, joint positions, and joint velocities. Experimental results for a single-link arm are also presented and compared with the case when the slow control is designed based on the rigid-body model of the manipulator.

In this paper, an index of optimization is proposed for improving the dynamic behavior of structurally flexible manipulators. The improvement is achieved through an optimization scheme where a cost function associated with the lowest natural flexible mode and an index defined as modal accessibility, is optimized subject to certain constraints. The formulation assumes a singularly perturbed model of the flexible-link system as this system possesses two-time scale properties. To illustrate the approach, the design of a two-link, non-uniform, planar, flexible manipulator is considered which results in improved performance characteristics as compared to a uniform manipulator.

In this paper a nonlinear control strategy for tip position trajectory tracking of a class of structurally flexible multi-link manipulators is developed. Using the concept of integral manifolds and singular perturbation theory, the full-order flexible system is decomposed into corrected slow and fast subsystems. The tip position vector is similarly partitioned into corrected slow and fast outputs. To ensure an asymptotic tracking capability, the corrected slow subsystem is augmented by a dynamical controller in such a way that the resulting closed-loop zero-dynamics are linear and asymptotically stable. The tracking problem is then re-defined as tracking the slow output and stabilizing the corrected fast subsystem by using dynamic output feedback. A major advantage of the proposed strategy is that the only measurements required are the tip positions, joint positions, and joint velocities.

This paper focuses on the tip-position control of a single flexible link which rotates in the horizontal plane. The dynamic model is derived using a Lagrangian assumed modes method based on Euler-Bernoulli beam theory. The model is then linearized about an operating point. An output feedback control strategy that uses the principle of transmission zero assignment achieves tracking for this nonminimum phase linear time-invariant system. The control strategy consists of two parts. The first part is an inner (stabilizing) control loop that incorporates a feedthrough term to assign the system's transmission zeros at desired locations in the complex plane, and a feedback term to move the system's poles to appropriate positions in the left-half plane. The second part is a feedback servo loop that allows tracking of the desired trajectory. The controller is implemented on an experimental test-bed. The performance is compared with that of a second controller based on pole placement state feedback.

The problem of controlling robot manipulators with flexible joints is considered. A reduced-order flexible-joint model based on a singular perturbation formulation of the manipulator equations of motion is used. The concept of an integral manifold is utilized to construct the dynamics of the slow subsystem. A fast subsystem is constructed to represent the dynamics of the elastic forces at the joints. A composite adaptive control scheme is developed with special attention to stability and robustness of the controller. The proposed controller is based on online identification of the manipulator parameters and takes into account the effect of a class of unmodeled dynamics, identification errors, and parameter variations. Stability analysis of the resulting closed-loop full-order system is presented. To show the capability of the proposed algorithm, an example of a two-link flexible-joint manipulator is considered. Simulation results are given to illustrate the applicability of the proposed control scheme.>

The problem of designing robust adaptive control strategies for a flexible-joint robot manipulator is considered. By utilizing the concept of integral manifolds, a corrected reduced-order model of the flexible system is obtained. Adaptive control schemes for the corrected reduced-order model are then developed that otherwise would have been difficult to obtain for the full-order flexible system due to ill conditioning and the curse of dimensionality. Zeroth-order singular perturbation results are generalized to include corrected adaptive control schemes. The result is a robust adaptive control law that takes both parametric and dynamic uncertainties into account. Two common adaptive control strategies, the adaptive inverse dynamics scheme and the Slotine and Li (1987) scheme, are examined. Numerical simulations for a two-link flexible-joint manipulator are carried out to illustrate the potential and advantages of the adaptive control schemes as compared to these two techniques.>

The problem of designing a robust adaptive control strategy for flexible joint robot manipulators is considered. By utilising the concept of integral manifolds, reduced-order models of the flexible system are obtained. This makes it possible to develop adaptive control schemes for the flexible full-order system at the reduced-order level, which otherwise would be difficult due to ill-conditioning and the curse of dimensionality. Two common adaptive control strategies are examined: (i) the adaptive inverse dynamics and (ii) the Slotine and Li algorithm. It is shown how these standard rigid adaptive control techniques can be generalized and improved for successful application to flexible joint manipulators. The result is robust adaptive control laws which take both parametric and dynamic uncertainties into account. Numerical simulations for a two-link flexible joint manipulator illustrate the potential and advantages of the new adaptive schemes as compare to the two standard algorithms in the literature.>

Various neural models are considered for solving the robot forward kinematics problem. It is demonstrated that a three-layer backpropagation network is capable of learning the forward kinematics of a rigid-link, open-chain manipulator without knowledge of the manipulator's kinematic structure. Simulation results show that, by properly training such a network, it is possible to model the forward kinematics with an acceptable degree of accuracy. However, it is also shown that, if information about the kinematic structure of a manipulator is available, a functional link network gives, by far, the most accurate results

A 2-D model reduction using singular perturbation methodology is presented. The strong-weak coupling effects that lead to eigenvalue clustering of A/sub 1/ and A/sub 4/ matrices are utilized to derive structure transformations for fast-slow decomposition. This in conjunction with an aggregation procedure given allows the development of reduced-order 2-D slow and fast subsystems that capture the essential behavior of the full-order system as evidenced by close examination of the relevant characteristic polynomials.>

A robust nonlinear adaptive control strategy is proposed to stabilize a robot manipulator subject to parametric and dynamic uncertainties. This is accomplished by utilizing state feedback and applying a control space transformation to a reduced-order model without parametric uncertainty. Under appropriate assumptions on the nature of the reduced-order uncertainties, an adaptive control strategy that compensates the nonlinear uncertainties is designed for the reduced-order feedback linearized system. An advantage of the proposed technique is that it requires external (feedback) linearization and update laws to be derived only from the reduced-order model, although the control strategy is applied to the full system. Numerical results for a single-link uncertain flexible-joint manipulator are included to illustrate the implementation and testing of the proposed method.>

The control problem for robot manipulators with flexible joints is considered. The results are based on a recently developed singular perturbation formulation of the manipulator equations of motion where the singular perturbation parameter µ is the inverse of the joint stiffness. For this class of systems it is known that the reduced-order model corresponding to the mechanical system under the assumption of perfect rigidity is globally linearizable via nonlinear static-state feedback, but that the full-order flexible system is not, in general, linearizable in this manner. The concept of integral manifold is utilized to represent the dynamics of the slow subsystem. The slow subsystem reduces to the rigid model as the perturbation parameter µ tends to zero. It is shown that linearizability of the rigid model implies linearizability of the flexible system restricted to the integral manifold. Based on a power series expansion of the integral manifold around µ = 0, it is shown how to approximate the feedback linearizing control to any order in µ. The result is then an approximate feedback linearization which, assuming stability of the fast variables, linearizes the system for all practical purposes.

In this paper we examine a recently developed singular perturbation formulation of the equations of motion for a robot manipulator with flexible joints, where the fast variables are the elastic forces at the joints and their time derivatives. The concept of an invariant manifold is utilized to represent the dynamics of the slow subsystem. The dynamics of the system restricted to this manifold reduce to the usual rigid body dynamics as the perturbation parameter ε tends to zero. Based on a power series expansion of the exact manifold around ε = 0, higher order corrections of the manifold are obtained. This leads to reduced order models of the full system which may prove more useful for control system design than either the full model or the rigid model. The case of a single link with joint flexibility is worked out in detail.