Iterative Learning Approach Research Articles

Tracking control of nonlinear lumped mechanical continuous-time systems: A model-based iterative learning approach

This paper presents a nonlinear model-based iterative learning control procedure to achieve accurate tracking control for nonlinear lumped mechanical continuous-time systems. The model structure used in this iterative learning control procedure is new and combines a linear state space model and a nonlinear feature space transformation. An intuitive two-step iterative algorithm to identify the model parameters is presented. It alternates between the estimation of the linear and the nonlinear model part. It is assumed that besides the input and output signals also the full state vector of the system is available for identification. A measurement and signal processing procedure to estimate these signals for lumped mechanical systems is presented. The iterative learning control procedure relies on the calculation of the input that generates a given model output, so-called offline model inversion. A new offline nonlinear model inversion method for continuous-time, nonlinear time-invariant, state space models based on Newton's method is presented and applied to the new model structure. This model inversion method is not restricted to minimum phase models. It requires only calculation of the first order derivatives of the state space model and is applicable to multivariable models. For periodic reference signals the method yields a compact implementation in the frequency domain. Moreover it is shown that a bandwidth can be specified up to which learning is allowed when using this inversion method in the iterative learning control procedure. Experimental results for a nonlinear single-input-single-output system corresponding to a quarter car on a hydraulic test rig are presented. It is shown that the new nonlinear approach outperforms the linear iterative learning control approach which is currently used in the automotive industry on durability test rigs.

SVM-HUSTLE—an iterative semi-supervised machine learning approach for pairwise protein remote homology detection

As the amount of biological sequence data continues to grow exponentially we face the increasing challenge of assigning function to this enormous molecular 'parts list'. The most popular approaches to this challenge make use of the simplifying assumption that similar functional molecules, or proteins, sometimes have similar composition, or sequence. However, these algorithms often fail to identify remote homologs (proteins with similar function but dissimilar sequence) which often are a significant fraction of the total homolog collection for a given sequence. We introduce a Support Vector Machine (SVM)-based tool to detect homology using semi-supervised iterative learning (SVM-HUSTLE) that identifies significantly more remote homologs than current state-of-the-art sequence or cluster-based methods. As opposed to building profiles or position specific scoring matrices, SVM-HUSTLE builds an SVM classifier for a query sequence by training on a collection of representative high-confidence training sets, recruits additional sequences and assigns a statistical measure of homology between a pair of sequences. SVM-HUSTLE combines principles of semi-supervised learning theory with statistical sampling to create many concurrent classifiers to iteratively detect and refine, on-the-fly, patterns indicating homology. When compared against existing methods for identifying protein homologs (BLAST, PSI-BLAST, COMPASS, PROF_SIM, RANKPROP and their variants) on two different benchmark datasets SVM-HUSTLE significantly outperforms each of the above methods using the most stringent ROC(1) statistic with P-values less than 1e-20. SVM-HUSTLE also yields results comparable to HHSearch but at a substantially reduced computational cost since we do not require the construction of HMMs. The software executable to run SVM-HUSTLE can be downloaded from http://www.sysbio.org/sysbio/networkbio/svm_hustle

An iterative learning approach for density control of freeway traffic flow via ramp metering

In this work, we apply the iterative learning control approach to address the traffic density control problem in a macroscopic level freeway environment with ramp metering. The traffic density control problem is first formulated into an output tracking and disturbance rejection problem. Through rigorous analysis, it is shown that the iterative learning control method can effectively deal with this class of control problem and greatly improve the traffic response. Next, the iterative learning control is combined with error feedback in a complementary modular manner to achieve the output tracking and system robustness. The effectiveness of the new approach is further verified through case studies with intensive simulations.

Phenomenography: a conceptual framework for information literacy education

PurposeThe purpose of this paper is to explore the adoption of a phenomenographic conceptual framework to investigate learning from the perspective of the learner, with the aim of reflecting on the features that this approach shares with information literacy education in general, and with the relational model in particular.Design/methodology/approachThe study offers an analysis of phenomenographic research on learning undertaken by Marton, which is further elaborated by examples of collaborative work by Marton and Booth, as well as by Fazey and Marton. The relationship between understanding and learning, promoted by this perspective, is explored in this paper to illustrate its impact on retention and transfer of the learning process. This is compared with the iterative and independent learning approaches promoted by information literacy education, and specific examples are used to illustrate the pedagogical overlap between phenomenography and information literacy. In addition, the paper examines the relational approach of information literacy promoted by the individual and collective works of Bruce, Lupton, and Edwards to demonstrate how the person‐world relation, advocated by phenomenography, is used to examine the learner‐information relationship promoted by the work of these authors.FindingsThe paper reflects on the potential impact that phenomenography and the relational perspectives have on pedagogical practices in Higher Education. In particular, it aims to demonstrate how the relational approach, together with the learn‐how‐to‐learn ethos of information literacy, is fundamental in promoting a framework for lifelong learning that leads to the empowering of the learner through an iterative cycle of reflection and practice, i.e. what phenomenography defines as variation in practice to foster the ownership of learning.Originality/valueIn line with the person‐world relation, the paper explores the relationship between learners and information by outlining its internal/subjective and external/objective dynamics. Claims that the learner's ability to reflect on these dynamics enhances his or her independent learning attitude are explored in the light of current phenomenographic and information literacy research.

Freeway Traffic Control Using Iterative Learning Control-Based Ramp Metering and Speed Signaling

In this paper, an iterative learning approach for the freeway density control under ramp metering and speed regulation is developed in a macroscopic level traffic environment. Rigorous analyses show that the proposed learning control schemes guarantee the asymptotic convergence of the traffic density to the desired one. The two major features of the learning-based density control are: 1) less prior modeling knowledge required in the control system design and 2) the ability to reject exogenous traffic perturbations. The control schemes are applied to a freeway model, and simulation results confirm the efficacy of the proposed approach

Local identification of prototypes for genetic learning of accurate TSK fuzzy rule-based systems

This work presents the use of local fuzzy prototypes as a new idea to obtain accurate local semantics-based Takagi–Sugeno–Kang (TSK) rules. This allow us to start from prototypes considering the interaction between input and output variables and taking into account the fuzzy nature of the TSK rules. To do so, a two-stage evolutionary algorithm based on MOGUL (a methodology to obtain Genetic Fuzzy Rule-Based Systems under the Iterative Rule Learning approach) has been developed to consider the interaction between input and output variables. The first stage performs a local identification of prototypes to obtain a set of initial local semantics-based TSK rules, following the Iterative Rule Learning approach and based on an evolutionary generation process within MOGUL (taking as a base some initial linguistic fuzzy partitions). Because this generation method induces competition among the fuzzy rules, a postprocessing stage to improve the global system performance is needed. Two different processes are considered at this stage, a genetic niching-based selection process to remove redundant rules and a genetic tuning process to refine the fuzzy model parameters. The proposal has been tested with two real-world problems, achieving good results. © 2007 Wiley Periodicals, Inc. Int J Int Syst 22: 909–941, 2007.

Design of a fuzzy controller in mobile robotics using genetic algorithms

The design of fuzzy controllers for the implementation of behaviors in mobile robotics is a complex and highly time-consuming task. The use of machine learning techniques, such as evolutionary algorithms or artificial neural networks for the learning of these controllers allows to automate the design process. In this paper, the automated design of a fuzzy controller using genetic algorithms for the implementation of the wall-following behavior in a mobile robot is described. The algorithm is based on the Iterative Rule Learning (IRL) approach, and a parameter (@d) is defined with the aim of selecting the relation between the number of rules and the quality and accuracy of the controller. The designer has to define the universe of discourse and the precision of each variable, and also the scoring function. No restrictions are placed neither in the number of linguistic labels nor in the values that define the membership functions.

Iterative Learning Approach Toward Closed-Loop Automatic Tuning of PID Controllers

This paper proposes a new method for closed-loop automatic tuning of a proportional−integral−derivative (PID) controller based on a new iterative learning control (ILC) approach. The proposed approach is applicable to process control applications where there is usually a time-delay/lag phenomenon and where nonrepetitive step changes in the reference signal are far more common than repetitive ones assumed in most literature on ILC. The method does not require the control loop to be detached for tuning, but it requires the input of a periodic reference signal which can be specified by the user or derived from a relay feedback experiment. A modified ILC scheme iteratively changes the control signal by adjusting the reference signal only. The learning gain can be selected to satisfy a necessary and sufficient condition derived in the paper, based on the information available from the oscillations induced. Once a satisfactory performance is achieved, the PID controller is then tuned by fitting the controller to yield a fitting input and output characteristic of the ILC component. Simulation and experimental results are furnished to illustrate the effectiveness of the proposed tuning method.

New trends in fuzzy modeling. part II: applications

Nowadays, one of the most important areas of application of fuzzy set theory are fuzzy rule-based systems. These kinds of systems constitute an extension of classical rule-based systems, because they deal with “IF-THEN” rules whose antecedents and consequents are composed of fuzzy logic statements instead of classical logic. They have been successfully applied to a wide range of problems in different domains for which uncertainty and vagueness emerge in different ways. The current special issues on “New trends in fuzzy modeling (Part I: Novel approaches and Part II: Applications) collect several contributions by experts in the topic that go beyond on the design of fuzzy systems with a balance between new developments and applications. This second issue is devoted to different representative applications of fuzzy systems in different areas, analysing the benefits of using them in these domains. The application domains are power robotic, market, transfer passenger movements and the finding of fuzzy sequential patterns from quantitative data for real-world applications. The papers in this issue can be classified according to their application domain. The first three papers in this issue belong to the first group, robotic. In the first paper, entitled “Hierarchical fuzzy rule based systems using an information theoretic approach”, A. Waldock et al. propose a novel anytime algorithm for the construction of a Hierarchical Fuzzy Based System using an information theoretic approach to specialise rules that do not effectively model the decision space. The proposed approach is demonstrated on a mobile robot application in simulation. M. Mucientes et al. propose an evolutionary fuzzy system based on the Iterative Rule Learning approach in the paper “Evolutionary learning of a fuzzy controller for wall-follow-

Evolutionary learning of a fuzzy controller for wall-following behavior in mobile robotics

The design of fuzzy controllers for the implementation of behaviors in mobile robotics is a complex and highly time-consuming task. The use of machine learning techniques such as evolutionary algorithms or artificial neural networks for the learning of these controllers allows to automate the design process. In this paper, the automated design of a fuzzy controller using genetic algorithms for the implementation of the wall-following behavior in a mobile robot is described. The algorithm is based on the iterative rule learning approach, and is characterized by three main points. First, learning has no restrictions neither in the number of membership functions, nor in their values. In the second place, the training set is composed of a set of examples uniformly distributed along the universe of discourse of the variables. This warrantees that the quality of the learned behavior does not depend on the environment, and also that the robot will be capable to face different situations. Finally, the trade off between the number of rules and the quality/accuracy of the controller can be adjusted selecting the value of a parameter. Once the knowledge base has been learned, a process for its reduction and tuning is applied, increasing the cooperation between rules and reducing its number.

An iterative learning approach to compensation for the servo track writing error in high track density disk drives

The servo tracks of hard disk drives are written at the time of manufacture with the equipment of servo track writer. The disk vibrations or head fluctuations during servo track writing process give rise to servo track writing error. The servo track writing error may cause some critical errors during data writing operation. In this paper, we propose a new correction algorithm for the servo track writing error based on the iterative learning control technique. The estimate of the servo track writing error is constructed from the position error signal and updated iteratively at each disk rotation. Then, the estimate of servo track writing error is used to correct the position error signal in a feed-forward manner. Our correction algorithm is robust to system model uncertainties, computationally quite simple, and has fast convergence rate. Furthermore, we give a rigorous analysis for the convergence of our correction algorithm. In order to demonstrate the practical use of our work, we present some experimental results using a commercially available hard disk drive.

Evolutionary approaches to fuzzy modelling for classification

An overview of the application of evolutionary computation to fuzzy knowledge discovery is presented. This is set in one of two contexts: overcoming the knowledge acquisition bottleneck in the development of intelligent reasoning systems, and in the data mining of databases where the aim is the discovery of new knowledge. The different strategies utilizing evolutionary algorithms for knowledge acquisition are abstracted from the work reviewed. The simplest strategy runs an evolutionary algorithm once, while the iterative rule learning approach runs several evolutionary algorithms in succession, with the output from each considered a partial solution. Ensembles are formed by combining several classifiers generated by evolutionary techniques, while co-evolution is often used for evolving rule bases and associated membership functions simultaneously. The associated strengths and limitations of these induction strategies are compared and discussed. Ways in which evolutionary techniques have been adapted to satisfy the common evaluation criteria of the induced knowledge—classification accuracy, comprehensibility and novelty value—are also considered. The review concludes by highlighting common limitations of the experimental methodology used and indicating ways of resolving them.

Combining boosting and evolutionary algorithms for learning of fuzzy classification rules

This paper presents a novel boosting algorithm for genetic learning of fuzzy classification rules. The method is based on the iterative rule learning approach to fuzzy rule base system design. The fuzzy rule base is generated in an incremental fashion, in that the evolutionary algorithm optimizes one fuzzy classifier rule at a time. The boosting mechanism reduces the weight of those training instances that are classified correctly by the new rule. Therefore, the next rule generation cycle focuses on fuzzy rules that account for the currently uncovered or misclassified instances. The weight of a fuzzy rule reflects the relative strength the boosting algorithm assigns to the rule class when it aggregates the casted votes. The approach is compared with other classification algorithms for a number problem sets from the UCI repository.

An Iterative Learning-Based Modulation Scheme for Torque Control in Switched Reluctance Motors

This paper deals with an iterative learning approach for modulating the desired torque profile so as to obtain ripple-free torque in switched reluctance motors. Because of the highly nonlinear relation between torque, current, and rotor position for this motor, it is not possible to obtain a closed-form mathematical expression for current as a function of torque and rotor position. Thus, the current waveforms are conventionally computed by using the linear torque model of the motor, and it is well known that such a scheme results in high torque ripple. In this paper, a novel method is proposed to minimize the ripple. In this new scheme, the current is still computed using the linear torque model, but the value of the torque used for this is not the desired (specified) torque, but rather a modulated-desired torque that is obtained by repeated corrections to the desired torque from iteration to iteration. The conventional rectangular pulse profile is taken as the initial current waveform. The method requires much less a priori knowledge of the magnetic characteristics of the motor. The algorithms have been formulated for both one-phase-on and two-phase-on schemes, for a four-phase switched reluctance motor, in the light of the principles behind iterative learning. Based on the observations from the simulation results of these schemes, a modified scheme has been proposed by incorporating a suitable commutation process, often called torque sharing functions, in order to generate reasonably smooth current waveforms for the ease of tracking by the stator circuit of the motor. The performances of all the proposed schemes have been verified by computer simulation.

Pole assignment for stochastic systems with unknown coefficients

This paper solves the exact pole assignment problem for the single-input stochastic systems with unknown coefficients under the controllability assumption which is necessary and sufficient for the arbitrary pole assignment for systems with known coefficients. The system noise is required to be mutually independent with zero mean and bounded second moment. Two approaches to solving the problem are proposed: One is the iterative learning approach which can be applied when the state at a fixed time can be repeatedly observed with different feedback gains; the other is the adaptive control approach which works when the trajectories satisfy a nondegeneracy condition. Both methods are essentially based on stochastic approximation, and the feedback gains are recursively given without invoking the certainty-equivalency-principle.

MOGUL: A methodology to obtain genetic fuzzy rule-based systems under the iterative rule learning approach

The main aim of this paper is to present MOGUL, a Methodology to Obtain Genetic fuzzy rule-based systems Under the iterative rule Learning approach. MOGUL will consist of some design guidelines that allow us to obtain different genetic fuzzy rule-based systems, i.e., evolutionary algorithm-based processes to automatically design fuzzy rule-based systems by learning and/or tuning the fuzzy rule base, following the same generic structure and able to cope with problems of a different nature. A specific evolutionary learning process obtained from the paradigm proposed to design unconstrained approximate Mamdani-type fuzzy rule-based systems will be introduced, and its accuracy in the solving of a real-world electrical engineering problem will be analyzed. ©1999 John Wiley & Sons, Inc.

MOGUL: A methodology to obtain genetic fuzzy rule‐based systems under the iterative rule learning approach

The main aim of this paper is to present MOGUL, a Methodology to Obtain Genetic fuzzy rule-based systems Under the iterative rule Learning approach. MOGUL will consist of some design guidelines that allow us to obtain different genetic fuzzy rule-based systems, i.e., evolutionary algorithm-based processes to automatically design fuzzy rule-based systems by learning and/or tuning the fuzzy rule base, following the same generic structure and able to cope with problems of a different nature. A specific evolutionary learning process obtained from the paradigm proposed to design unconstrained approximate Mamdani-type fuzzy rule-based systems will be introduced, and its accuracy in the solving of a real-world electrical engineering problem will be analyzed. ©1999 John Wiley & Sons, Inc.

Application of iterative learning for constant torque control of switched reluctance motors

Abstract The paper deals with the investigations on an iterative learning approach to determine the desired current waveforms for switched reluctance motors, which give rise to ripple-free torque. The current waveforms are generated by repeated corrections from iteration to iteration starting from the conventional rectangular pulse profile as the initial waveform. The scheme requires much less a priori knowledge of the magnetic characteristics of the motor. The algorithms have been formulated for both one-phase-on and two-phase-on schemes, for a four-phase switched reluctance motor, in the light of the principles behind iterative learning. Based on the observations from the test results of these schemes, a modified scheme has been proposed by incorporating a suitable commutation process, often called as torque sharing functions, in order to generate reasonably smooth current waveforms for the ease of tracking by the stator circuit of the motor. The performances of all the proposed schemes have been verified by computer simulation.

Determination of current waveforms for torque ripple minimisation in switched reluctance motors using iterative learning: an investigation

The paper deals with the investigations on an iterative learning approach to determine the desired current waveforms for switched reluctance motors, which give rise to ripple-free torque. The current waveforms are generated by repeated corrections from iteration to iteration starting from the conventional rectangular pulse profile as the initial waveform. The scheme requires much less a priori knowledge of the magnetic characteristics of the motor. The algorithms have been formulated for both one-phase-on and two-phase-on schemes, for a four-phase switched reluctance motor, in the light of the principles behind iterative learning. Based on the observations from the simulation results of these schemes, a modified scheme has been proposed by incorporating a suitable commutation process, often called torque sharing functions, in order to generate reasonably smooth current waveforms for the ease of tracking by the stator circuit of the motor. The performances of all the proposed schemes have been verified by computer simulation.

An Iterative Learning Approach to Torque Ripple Minimization in Switched Reluctance Motors

Switched reluctance motors are rapidly becoming popular because of their simple construction, ease of maintenance, low cost and high efficiency. However, because of its highly nonlinear nature of torque production mechanism, high torque ripple is generated with the conventional rectangular pulse excitation scheme. This paper deals with the preliminary investigations on the use of iterative learning to determine the current waveforms so as to minimize the torque ripple. The scheme requires much less a priori knowledge of the magnetic characteristics of the motor. The strategies have been formulated for both one-phase-on and two-phase-on schemes. The performance is verified with a four-phase switched reluctance motor by computer simulation.