Editorial for the Special Issue on Learning‐based Adaptive Control: Theory and Applications

The Special Issue presents results of current research on learning‐based adaptive methods, merging together model‐based and data‐driven adaptive approaches. The special issue contains two main types of contributions. The first type of papers presents new theoretical developments for learning‐based adaptive algorithms, while the second type focuses on challenging practical applications ranging from UAVs, and autonomous vehicles, to heating and ventilation systems. These papers are compiled in a special issue of the journal. To access all of the papers please follow the following link ( https://onlinelibrary.wiley.com/toc/10991115/2019/33/2 ).     

Non‐linear adaptive control of a heat exchanger

In this contribution, two methods for adaptation of non‐linear adaptive controllers are presented and compared, namely the data‐driven and the knowledge‐based adaptation. A dynamic Takagi–Sugeno fuzzy model is utilized to model the non‐linear process behaviour. Based on this model, a non‐linear predictive controller is designed to control the process. In the presence of time‐variant process behaviour and changing unmodelled disturbances, high control performance can be achieved by performing an on‐line adaptation of the fuzzy model. First, a local weighted recursive least‐squares algorithm is used for adaptation. It exploits the local linearity of the Takagi–Sugeno fuzzy model. In the second approach, process knowledge that is obtained from theoretical insights is utilized to design a knowledge‐based adaptation strategy. Both approaches are compared and their effectiveness and real‐world applicability are demonstrated by application to temperature control of a heat exchanger. Copyright © 2001 John Wiley & Sons, Ltd.     

Output‐feedback model‐reference adaptive calibration for map‐based anti‐jerk control of electromechanical automotive clutches

It is well known that the map‐based control can reduce the computational burden of the automotive on‐board controller. This paper proposes an output‐feedback model‐reference adaptive control algorithm to calibrate the map‐based anti‐jerk controller for electromechanical clutch engagement. The algorithm can be used to adaptively construct a data‐driven fuzzy rule base without resorting to manual tuning, so that it can overcome the problem of conventional knowledge‐based fuzzy logic design, which involves strenuous parameter‐tuning work in the construction of calibration maps. To accurately define the consequent of each fuzzy rule for anti‐jerk control, an output feedback law for computing the reference trajectory of clutch engagement is developed to eliminate the discontinuous slip‐stick transition, whereas an adaptive controller is designed to track the reference trajectory and compensate the nonlinearity. The convergence of the proposed output‐feedback model‐reference adaptive control algorithm is analyzed. Simulation results indicate that the proposed method can successfully reduce the excessive vehicle jerk and frictional energy dissipation during clutch engagement as compared with the conventional knowledge‐based fuzzy logic controller without fine tuning.     

Learning‐based iterative modular adaptive control for nonlinear systems

In this paper, we study the problem of adaptive trajectory tracking control for a class of nonlinear systems with structured parametric uncertainties. We propose to use an iterative modular approach: we first design a robust nonlinear state feedback that renders the closed‐loop input‐to‐state stable (ISS). Here, the input is considered to be the estimation error of the uncertain parameters, and the state is considered to be the closed‐loop output tracking error. Next, we propose an iterative adaptive algorithm, where we augment this robust ISS controller with an iterative data‐driven learning algorithm to estimate online the parametric uncertainties of the model. We implement this method with two different learning approaches. The first one is a data‐driven multiparametric extremum seeking method, which guarantees local convergence results, and the second is a Bayesian optimization‐based method called Gaussian Process Upper Confidence Bound, which guarantees global results in a compact search set. The combination of the ISS feedback and the data‐driven learning algorithms gives a learning‐based modular indirect adaptive controller. We show the efficiency of this approach on a two‐link robot manipulator numerical example.     

A comparison of model‐based and data‐driven controller tuning

In many industrial applications, finding a model from physical laws that is both simple and reliable for control design is a hard and time‐consuming undertaking. When a set of input/output measurements is available, one can derive the controller directly from data, without relying on the knowledge of the physics. In the scientific literature, two main approaches have been proposed for control system design from data. In the ‘model‐based’ approach, a model of the system is first derived from data and then a controller is computed‐based on the model. In the ‘data‐driven’ approach, the controller is directly computed from data. In this work, the previous approaches are compared from a novel perspective. The main finding of the paper is that, although from the standard perspective of parameter variance analysis the model‐based approach is always statistically more efficient, the data‐driven controller might outperform the model‐based solution for what concerns the final control cost. Copyright © 2013 John Wiley & Sons, Ltd.     

A scheduling method considering register‐to‐register transfers

In high‐level synthesis, scheduling is an important stage which assigns each operation appearing in a data flow graph to a specific control step, whose results influence the design quality directly. This paper describes a scheduling approach for pipelined datapaths. Since few previous approaches estimate the interconnection cost between register (register‐to‐register cost), our approach introduces a datapath model with interconnection between registers across buses, and minimizes the total hardware cost including the register‐to‐register cost by force‐directed scheduling. © 1999 Scripta Technica, Electr Eng Jpn, 128(3): 63–71, 1999     

A discrete‐time indirect adaptive multiple‐model actuator failure compensation scheme

A new discrete‐time actuator failure compensation control scheme is developed, using a multiple‐model adaptive control approach which has the capacity to achieve faster and more accurate compensation of failure uncertainties. An individual adaptive system, for each possible failure pattern in a failure pattern set of interest for compensation, is designed using an indirect model reference adaptive control scheme for actuator failure compensation. A multiple‐model control switching mechanism for discrete‐time systems is set up by finding the minimal performance index to select the most appropriate control law. The performance indices are based on the adaptive estimation errors of individual parameterized systems with actuator failures. Simulation results from an aircraft flight control system example are presented to show the desired closed‐loop system stability and tracking performance despite the presence of uncertain actuator failures. Copyright © 2014 John Wiley & Sons, Ltd.     

Simple and accurate approaches to implement the complex trans‐conductance suited for time‐domain simulators for small‐signal and large‐signal table‐based models

This paper focuses on the implementation of table‐based models of high‐frequency transistors for time‐domain simulators at microwave and mm‐wave frequencies. In this frequency range, the channel is not capable of responding to the excitation instantaneously therefore, a delay‐time exists between the channel response and the channel excitation. This delay is represented by a complex trans‐conductance in terms of circuit elements. The high‐frequency models of transistors are required to have the implementation of complex trans‐conductance, where the complex part accounts mathematically for the delay‐time between the channel response and the channel excitation. This paper presents simple and accurate approaches to incorporate the complex trans‐conductance in both small‐signal and large‐signal table‐based models for time‐domain simulators (MOS‐AK International Meeting. Eindhoven, Netherlands, April 2008). Implementation approach for each model, small‐signal and large‐signal, is presented in separated sections. In the first step, the delay is realized by the introduction of an ideal transmission line between the channel excitation and the channel response. As transmission lines are not generally suitable for time‐domain simulations, a lumped element equivalent network is introduced in the second step. The latter approach is fully compatible with time‐domain simulators but frequency limitation, determined by the delay‐time value itself, is introduced. Then the implementation of the complex trans‐conductance in large‐signal model is introduced. In terms of large‐signal behavior, delay‐time is important to achieve a non‐quasi static model. Yet again there is limitation in terms of the frequency range that is determined by the delay value itself. The methodology is illustrated on the small‐signal and the large‐signal equivalent circuit of a Multi‐Fin MOSFET transistor. Simulations are carried out by Cadence Spectre and Agilent ADS simulators, and comparisons are carried out between the simulation results and the measurements. Copyright © 2010 John Wiley & Sons, Ltd.     

Data‐Driven Controller Tuning for Nonminimum Phase Plants with Stability Constraints

This paper addresses the model reference control problem, which is a typical control problem found in data‐driven controller tuning methods. For nonminimum phase plants, the unstable zeros of the plant should be included in the reference to avoid destabilization of the resulting closed‐loop system and improve tracking performance. First, we propose a data‐driven controller tuning method with closed‐loop stability taken into consideration and with the tuned controller parameters in the time domain. If the plant has unstable zero(s), the proposed method would not lead to destabilizing controller in the worst case. Closed‐loop stability is checked using linear inequalities described with input/output data. This contributes to reducing computation in the proposed method. Moreover, this paper proposes a data‐driven controller tuning method for nonminimum phase plants estimating the unstable zero(s) using a flexible reference model at each parameter update and reflecting them into the resulting reference model. The effectiveness of the proposed method is confirmed through numerical experiments.     

A semi‐adaptive control approach to closed‐loop medication infusion

This paper presents a semi‐adaptive control approach to closed‐loop medication infusion problems. The rationale underlying this approach is to design a controller that can adapt model parameters with a large impact on the model's fidelity while fixing the remaining parameters at nominal values. In this paper, a control‐oriented model for this purpose is derived via system identification and sensitivity analysis of a low‐order model capturing the direct dose‐response relationship Using the model thus derived, a model‐reference adaptive controller and a composite adaptive controller are designed and compared with each other. In‐silico simulation results using remifentanil's effect on respiratory rate as an example indicate that both controllers can regulate the output at commanded set points. Copyright © 2016 John Wiley & Sons, Ltd.     

Multi‐input multi‐output model‐free predictive control and its application to wastewater treatment

Model‐free predictive control is a data‐driven control method that directly computes the control input from massive input/output datasets. It does not require the mathematical models that are used in conventional model predictive control. It has recently been shown that the control offered by model‐free predictive control can be improved by the introduction of polynomial regression vectors containing the control input and measurement output. In this paper, we extend these findings to multi‐input multi‐output nonlinear systems and investigate the effectiveness of the approach through numerical simulations of a wastewater treatment process. © 2017 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.     

Development of mathematical model for signal‐dependent linear adaptive level‐based sampling

A mathematical model for linear adaptive level‐based sampling is developed in this paper. The proposed mathematical model provides a gradual change in quantization size and maintains time difference between two successive samples greater than or equal to loop delay of the system. This model is developed to minimize signal slope overload distortion as well as reduction in quantization noise. The proposed mathematical model is demonstrated in MATLAB simulation environment. Simulation results are compared with level‐based sampling and adaptive level‐based sampling based on bit‐level compression. Copyright © 2015 John Wiley & Sons, Ltd.     

Experience replay–based output feedback Q‐learning scheme for optimal output tracking control of discrete‐time linear systems

This paper focuses on solving the adaptive optimal tracking control problem for discrete‐time linear systems with unknown system dynamics using output feedback. A Q‐learning‐based optimal adaptive control scheme is presented to learn the feedback and feedforward control parameters of the optimal tracking control law. The optimal feedback parameters are learned using the proposed output feedback Q‐learning Bellman equation, whereas the estimation of the optimal feedforward control parameters is achieved using an adaptive algorithm that guarantees convergence to zero of the tracking error. The proposed method has the advantage that it is not affected by the exploration noise bias problem and does not require a discounting factor, relieving the two bottlenecks in the past works in achieving stability guarantee and optimal asymptotic tracking. Furthermore, the proposed scheme employs the experience replay technique for data‐driven learning, which is data efficient and relaxes the persistence of excitation requirement in learning the feedback control parameters. It is shown that the learned feedback control parameters converge to the optimal solution of the Riccati equation and the feedforward control parameters converge to the solution of the Sylvester equation. Simulation studies on two practical systems have been carried out to show the effectiveness of the proposed scheme.     

Robust control and data‐driven tuning of a hybrid integrator‐gain system with applications to wafer scanners

A hybrid integrator‐gain system is discussed that aims for improved low‐frequency disturbance rejection, while, at the same time, does not deteriorate overshoot and settling times when compared with a linear integrator. The hybrid integrator has similar phase advantages as the well‐known Clegg integrator but without inducing the discontinuous behavior resulting from resetting system state values. Optimal tuning of the controller parameters of the hybrid integrator is strongly influenced by machine‐specific properties and therefore favors a data‐driven optimization approach. However, as a time‐domain optimization algorithm can easily lead to nonrobust solutions in the sense of large peaking of the closed‐loop frequency response functions, frequency‐domain robustness constraints will be imposed. By means of an adaptive weighting filter design, the parameter updates are penalized upon violation of said robustness constraints. Posed in an unconstrained problem formulation, this is subsequently solved by applying a Gauss‐Newton–based parameter update scheme. Closed‐loop stability of the linear time‐invariant plant and controller in feedback connection with a hybrid integrator‐gain system element follows from a circle‐criterion‐like analysis, which is based on evaluating (measured) frequency response data. Measurement results obtained from an industrial wafer scanner demonstrate the effectiveness of the approach.     

Adaptive set‐point regulation of discrete‐time nonlinear systems

In this paper, adaptive set‐point regulation controllers for discrete‐time nonlinear systems are constructed. The system to be controlled is assumed to have a parametric uncertainty, and an excitation signal is used in order to obtain the parameter estimate. The proposed controller belongs to the category of indirect adaptive controllers, and its construction is based on the policy of calculating the control input rather than that of obtaining a control law. The proposed method solves the adaptive set‐point regulation problem under the assumption that the target state is reachable for each fixed parameter value. Additional feature of the proposed method is that Lyapunov‐like functions have not been used in the construction of the controllers. Copyright © 2015 John Wiley & Sons, Ltd.     

Accommodation of unknown actuator faults using output feedback‐based adaptive robust control

In this paper, we solve the problem of output tracking for linear uncertain systems in the presence of unknown actuator failures using discontinuous projection‐based output feedback adaptive robust control (ARC). The faulty actuators are characterized as unknown inputs stuck at unknown values experiencing bounded disturbance and actuators losing effectiveness at unknown instants of time. Many existing techniques to solve this problem use model reference adaptive control (MRAC), which may not be well suited for handling various disturbances and modeling errors inherent to any realistic system model. Robust control‐based fault‐tolerant schemes have guaranteed transient performance and are capable of dealing with modeling errors to certain degrees. But, the steady‐state tracking accuracy of robust controllers, e.g. sliding mode controller, is limited. In comparison, the backstepping‐based output feedback adaptive robust fault‐tolerant control (ARFTC) strategy presented here can effectively deal with such uncertainties and overcome the drawbacks of individual adaptive and robust controls. Comparative simulation studies are performed on a linearized Boeing 747 model, which shows the effectiveness of the proposed scheme. Copyright © 2011 John Wiley & Sons, Ltd.     

Improving Networks‐on‐Chip performability: A topology‐based approach

The performability metric is commonly used in Networks‐on‐Chip (NoC)‐based systems to represent their abilities to successfully complete specific tasks in finite time intervals. In this paper, we present a novel topology‐based performability model for NoC‐based systems. The model is used to evaluate the performability of NoC‐based systems at early design phases. A comparative study of nine commonly used network architectures is performed using the proposed model. The purpose of the study is to explore the impact of the network topology on the performability of NoC‐based systems. Using the output from this study, a new methodology is proposed to improve the performability of a given application at early design phases. In this methodology, a joint consideration of five design parameters (network topology, target application traffic distribution, mapping of processing elements, noise power, and voltage swing) is carried out. Using the proposed methodology, designers can select the optimal topology for a given application that maximizes system performability. The effectiveness of the proposed methodology in determining the optimal topology is verified by experimental work and validated through a case study of a video application. Copyright © 2010 John Wiley & Sons, Ltd.     

Multiple characteristic model‐based golden‐section adaptive control: Stability and optimization

The characteristic model‐based golden‐section adaptive control (CM‐GSAC) law has been developed for over 20 years in China with a broad range of applications in various fields. However, quite a few theoretical problems remain open despite its satisfying performance in practice. This paper revisits the stability of the CM‐GSAC from its very beginning and explores the underlying implications of the so‐called golden‐section parameter l₂≈0.618. The closed‐loop system, which consists of the CM and the GSAC, is a discrete time‐varying system, and its stability is discussed from three perspectives. First, attentions have been paid to select the optimal controller coefficients such that the closed‐loop system exhibits the best transient performance in the worst case. Second, efforts are made to improve the robustness in the presence of parameter estimation errors, which provide another choice when designing the adaptive controller. Finally, by measuring the slowly time‐varying nature in an explicit inequality form, a bridge is built between the instantaneous stability and the time‐varying stability. In order to relax the constraints on the parameter bounds of the CM, the GSAC is further extended to multiple CMs, which shows more satisfying tracking performance than that of the traditional multiple model adaptive control method. Copyright © 2014 John Wiley & Sons, Ltd.     

Coordinated semi‐adaptive closed‐loop control for infusion of two interacting medications

This paper presents a coordinated and semi‐adaptive closed‐loop control approach to the infusion of 2 interacting medications. The proposed approach consists of an upper‐level coordination controller and a lower‐level semi‐adaptive controller. The coordination controller recursively adjusts the reference targets based on the estimated dose‐response relationship of a patient to ensure that they can be achieved by the patient. The semi‐adaptive controller drives the patient outputs to the reference targets while estimating the patient's dose‐response relationship online. In this way, the controller is resilient to unachievable caregiver‐specified reference targets and responsive to the medication needs of individual patients. To establish the proposed approach, we developed the following: (1) a linear two‐input–two‐output dose‐response model; (2) a two‐input–two‐output semi‐adaptive controller to regulate the patient outputs while adapting high‐sensitivity parameters in the patient model; and (3) a coordination controller to adjust the reference targets that reconcile caregiver inputs and medication use. The proposed approach was applied to an example scenario in which cardiac output and respiratory rate are regulated via infusion of propofol and remifentanil in an in silico simulation setting. The results show that the coordinated semi‐adaptive control could (1) track achievable reference targets with consistent transient and steady‐state performance and (2) resiliently adjust the unachievable reference targets to achievable ones.     

Improved Inspection of Facilities for High‐Voltage Class Using Data Mining

The inspection of power supply facilities can now be conducted with high accuracy using remote monitoring technology. In contrast, it is difficult to install sensors at demand facilities because their scale and installation environment differ among customers. As a result, the demand facilities are inspected at fixed time intervals. In this paper, we propose condition‐based maintenance (CBM), which improves maintenance quality at demand facilities. The proposed method was developed using maintenance data from demand facilities, collected using time‐based maintenance, and we conduct the analysis primarily using failure data. We use data mining to analyze transaction data that we modeled on the basis of the maintenance data and to construct a “failure predictive model” that can predict the failure of facilities and its causes from the results of the analysis. By using the constructed model, we will be able to identify the objects requiring maintenance which may most likely lead to failures in the future, and this study can contribute to improvement of maintenance technologies for demand facilities using the proposed CBM.