When first introduced half a century ago, adaptive control was half accepted as useful and half rejected as useless in industrial systems, and has greatly evolved theoretically. Learning control, a related discipline, has also been widely studied, especially in robot control. Adaptive/learning control, which incorporates the two, has become trendy in Japan and elsewhere. New design methods, e.g., data-driven controllers and the machine learning based controllers, are also attracting attention.

This special issue, which focuses on adaptive/learning control, includes 18 contributions classified as follows:

• Closed-loop identification and controller redesign

• Adaptive output feedback control

• Data-driven control

• Multirate control

• Computational intelligence-based approaches

• Applications centering on electric motors, engine systems, hydraulic excavators, rotary cranes, etc. In addition, one review paper covers performance-driven control.

The theoretical study of adaptive/learning control has few actual applied examples in the form of real systems but is flourishing. Applied studies are expected to increasingly progress and adaptive/learning control theory holds big changes for industrial fields.

This study proposes a performance-driven control method that performs a “control performance assessment” and a “control system design” from a set of closed-loop data. The method assesses control performance based on the minimum variance control law from closed-loop data. It also calculates a control parameter that improves the control performance from the same closed-loop data by using the fictitious reference iterative tuning (FRIT) method. This method is characterized by not requiring any system model. The effectiveness of this method is verified through a numerical simulation and an application result to a temperature control unit.

This paper presents a new real-time parameter tuning in the data-driven framework. We focus on the tuning of state feedback gains to realize the desired performance of closed loop systems. For a real-time update tuning of this type of a controller, the notion of fictitious reference signal or the fictitious exogenous signal is utilized to generate the optimal gains in the real-time by using the measured past data. We also explain how the optimization can be realized as a recursive computation in real-time updates. Finally, an experiment is done to verify the effectiveness of the proposed method.

Increasing motor operating efficiency potentially reduces energy consumption and radiator requirements. In some applications, the motor’s operating efficiency may even be the most important factor to be considered. The model-based minimum total loss control (MTLC) we proposed could potentially improve motor efficiency by taking copper and iron loss into account based on the motor model, finding the minimum power point, and enabling the permanent magnet synchronous motor achieve optimal efficiency at different speeds. Copper loss under MTLC increased less than the maximum torque per ampere (MTPA), but iron loss under MTLC was much lower than MTPA, reducing total loss. MTLC reduced total loss by adjusting the relation between copper and iron loss using a demagnetization current. Analysis and simulation results show that our proposal improved the efficiency of the permanent magnet synchronous motor effectively, making it easy to implement in practical control with little calculation.

In the new design method we propose for a multirate output feedback control system, the hold interval of control input is longer than the sampling interval of plant output. In this system, unknown state variables are calculated using control input and plant output without observers. The multirate output feedback control system has been extended by introducing new design parameters that are designed independent of the calculation of the state variable. To our knowledge, however, no systematic design scheme has ever been proposed for design parameters in this case. In this study, quantization error is dealt with statistically and design parameters are decided to minimize quantization error.

We present a simple antisway control method for a rotary crane, whose load can move in a two-dimensional plane. In particular, we investigate the suppression of residual sway motion of a rotary crane with a boom that performs point-to-point motion. In the proposed method, we attempt to generate the trajectory of the boom using a combination of polynomial and cycloidal functions. The profile of the generated trajectory depends on the coefficients of the polynomial function. Thus, it is necessary to tune the coefficients to minimize the sway motion in the two-dimensional plane as much as possible. We adopt a particle swarm optimization algorithm, an evolutionary computation technique, to tune the coefficients and then obtain the optimal trajectory. By rotating the boom along the optimal trajectory, the two-dimensional residual sway motion is suppressed, i.e., an open-loop control is realized. The effectiveness and feasibility of the proposed control scheme is demonstrated via simulations and experiments.

According to a worldwide WHO survey, about one-third of people at the age of 65 or older experience at least one fall a year, which may result in a severe injury. Meanwhile, the population of the developed world is increasingly aging, and fall incidents can be therefore considered as a global problem. The causes of falls include the weakening of the tibialis anterior and gastrocnemius muscles that respectively play important roles in the dorsal and plantar flexion of the foot, and deterioration of the functions necessary to recover balance from perturbations during gait. Such dysfunctions are treated with rehabilitation provided by physical therapists and with special gait training in which the patient is subjected to perturbations. Although devices for rehabilitation and gait training have been developed, they are problematic since they only allow the ankle joint to move at a low number of degrees of freedom (DOF). In this study, we developed an ankle foot orthosis to provide six-DOF control of the ankle joint using a parallel link mechanism known as a Stewart platform. The Stewart platform construction makes it possible to provide six-DOF control. Since the ankle foot orthosis can be applied to walking, it can assist walking or gait training. In one of our prior studies, we proposed a force control method for the device, and verified its accuracy. In the present study, we improved the attachment method and introduced a pressure sensor to the previous version of the device to allow implementation of a new method that enables control adapted to the human gait. In addition, we conducted four experiments to verify whether it is possible to reproduce the physical therapist’s rehabilitation manipulations without limiting the ankle joint’s DOF, provide arbitrary walking assist action, and impart perturbations to the subject during gait. The first experiment verified the device’s accuracy in reproducing motion, the second confirmed the dispersion of the reproduced motion, the third assessed the walking-assist performance to prevent trips, and the fourth ascertained whether it is possible to make the subject lose balance by the imparted perturbation. The results demonstrated that the motions can be reproduced with high accuracy and with low dispersion and that the ankle joint motions can be controlled adaptively to fit the subject’s gait, suggesting the usefulness of the proposed device.

This paper deals with a combustion control system design problem of diesel engine. For a combustion model of diesel engine, which has been provided for on-board control, an adaptive control system based on an adaptive output feedback control with a simple adaptive feedforward input will be proposed based on the “almost strictly positive real-ness” (ASPR-ness) of the controlled system. A simple parallel feedforward compensator (PFC) design scheme, which renders the resulting augmented controlled system ASPR, will also be proposed based on basic combustion model properties to design stable adaptive control system for the diesel engine combustion control. The effectiveness of our proposal is confirmed through numerical simulation.

This paper proposes a design method of self-tuning generalized minimum variance control based on on-demand type feedback controller. A controller, such as generalized minimum variance control (GMVC), generalized predictive control (GPC) and so on, can be extended by using coprime factorization. Then new design parameter is introduced into the extended controller, and the parameter can re-design the characteristic of the extended controller, keeping the closed-loop characteristic that way. Although strong stability systems can be obtained by the extended controller in order to design safe systems, focusing on feedback signal, the extended controller can adjust the magnitude of the feedback signal. That is, the proposed controller can drive the magnitude of the feedback signal to zero if the control objective was achieved. In other words the feedback signal by the proposed method can appear on demand of achieving the control objective. Therefore this paper proposes on-demand type feedback controller using self-tuning GMVC for plant with uncertainty. A numerical example is shown in order to check the characteristic of the proposed method.

Wind energy use is being developed worldwide. Improving wind speed forecasting techniques has become important due to their economic impact on power system operation with increasing wind power penetration. Wind speed prediction is generally difficult due to wind’s intermittent nature, so many approaches have been proposed by researchers. The viability of these techniques has been verified, however, in only a certain few areas, rather than being evaluated quantitatively in many different locations. We use data from different parts of Japan for one-step-ahead prediction and applied different approaches at each point, which was then evaluated such as mean absolute error. We used the persistent model, the ARMA-GARCH model, the nonlinear autoregressive network with external input (NARX), the recurrent neural network (RNN), and support vector regression (SVR). Our results suggest that it is difficult to create the same model which minimizes error in all areas, confirming the need for individual predictors for individual regions.

This study investigated the effectiveness of a simple adaptive control system used for controlling the dynamics of the driving system of a wheeled mobile robot. This paper describes the design method of a speed control system and the gain adjustment procedure considering a practical application of the simple adaptive control system. Experiments were conducted on two different road conditions, and the results indicated that the rectilinear performance of the wheeled mobile robot improved with the use of the simple adaptive control system. In the case of uncertain road conditions, the simple adaptive control system reduced the deterioration of the control performance effectively.

The parameter-tuning method we discuss is for an Adaptive Nonlinear Model Predictive Controller (ANMPC). The MPC is optimization-based controller and decides control input to realize system output that tracks a reference trajectory through “optimal computation.” The reference trajectory is ideal trajectory of system output to converge on a desired value, i.e. controlled system performance depends on the reference trajectory. As a MPC controller which applies to the nonlinear systems, our group has already proposed an adaptive nonlinear MPC (ANMPC) for a tracking control problem of nonlinear two-link planar manipulators. This ANMPC uses a new reference trajectory having control parameters that must be tuned based on the desired controlled system’s responses and properties. To reduce troublesome parameter tuning, we propose new parameter-tuning method for ANMPC by a quantitative analysis of the relationship between a system’s behavior and ANMPC parameters. Numerically simulating the two-link nonlinear manipulator’s tracking control under various conditions demonstrates that proposed tuning method tunes the ANMPC effectively.

This paper discusses a design method for a multirate system including integrators, where the update interval of the control input is shorter than the sampling interval of the plant output. In such a multirate control system, intersample output might oscillate between sampled outputs in the steady state even if the sampled output converges to the reference input. This is because the control input can be updated between the sampled outputs. In a conventional method, a predesigned control law is extended such that the steady-state ripples are eliminated independent of a discrete-time response. However, the conventional method is invalid when integrators are included in a controlled plant. In this study, a difference operation in discrete time is used to address this issue. Moreover, the transient-state intersample response is improved independent of a pre-designed discrete-time response.

Prefilters for Virtual Reference Feedback Tuning (VRFT) in the time domain are realized by the least-squares method with step reference response data acquired from the closed-loop system to be tuned and are used to obtain controller parameters for making closed-loop systems as close as possible to the desired reference model. The usefulness of this proposal is shown in numerical examples in which fifth-order and PID controllers for a flexible transmission system are tuned by VRFT with prefilters.

In recent years, technology that includes informatization and automation has been introduced in the construction field. On the other hand, those field still require human operation technology based on experience and skills because various environmental conditions vary from hour to hour. Seasoned technicians have made such operation technology effective at various sites and established skillful techniques. However, the decreasing number and aging of skilled technicians are a social issue, making the skill tradition and development of younger technicians difficult at operation sites that require skillful techniques. This study assumed that the operation of machines by an operator was synonymous with the control of systems by a controller; human operation techniques were considered from the viewpoint of control engineering by regarding an operator as a controller. The control system used to represent the operator consisted of a proportional-integral-derivative (PID) controller and a cerebellar model articulation controller (CMAC) that adjusted the PID gains. A CMAC which is a type of neural network learns human skills as variations in the PID gains and expresses them based on the variations. This study applies the proposed method to a hydraulic excavator swing operation to evaluate skills. Moreover, the difference in the operation skills for the excavator is clarified by obtaining operation data for skilled and younger technicians and examining the variation tendency of PID gains.

This paper presents a data-oriented technique for designing a proportional-integral-derivative (PID) controller based on a generalized predictive control law for linear unknown systems. In several control design approaches, a model-based control theory, which requires accurate modeling and identification of the plant, is used to calculate the control parameters. However, in higher-order systems and/or systems with an unknown time delay such as chemical industries and thermal industries, it is difficult to model or identify the plant accurately. Over the last decade, data-oriented techniques in which the online or offline data are utilized have been attracting considerable attention. Designing the controllers for unknown plants based on only the input/output data is the main feature of this technique. In this study, controller parameters are first obtained by using a generalized predictive control law with the data-oriented technique, and are converted to PID parameters from the practical point of view. The proposed method is validated experimentally using a real injection-molding machine. The results demonstrate the efficiency of the proposed method.

In recent years, due to the mass retirement of skilled workers, loss of expertise has emerged as a problem in Japan. Meanwhile, the performance of computer hardware has been drastically improving. Skill-based PID controllers utilizing a database have been proposed as a potential solution to this problem. However, these controllers may not respond to multiple demands of control performance from users because the controller was not considered in the evaluation from the users. As a solution to this problem, an expert controller based on a skilled worker’s operating information with control performance assessments has been proposed. According to the method, I/O data, PID parameters that are estimated using the operating data and evaluation values of the skill of the skilled worker are stored in the database. From this, information vectors with high scores are selected, and a local PID controller is designed in response to the user’s requirements. In the conventional research, the least squares method is applied for estimating the PID parameters from the operating data of the skilled worker, and there are no restrictions on their values. This risks a loss of physical meanings of PID parameters in the case that they have negative values. In this research, an expert controller using particle swarm optimization (PSO) is proposed. In this method, data obtained by human control of a control simulator constructed on a computer is used to estimate human’s skill as PID parameters. Moreover, providing restrictions for the estimation of PID parameters enables them to preserve their physical meanings. In this research, the effectiveness of the proposed expert controller is verified using a control simulator.

Virtual Reference Feedback Tuning (VRFT), proposed by Campi et al., is an effective data-driven tuning method used in feedback controllers because the desired parameters implemented in the controller are obtained by using only one-shot experiment data. In this paper, we apply VRFT to cascade control systems. We also discuss the meaning of the cost function to be minimized. A numerical example is demonstrated to show an effectiveness and validity of our proposed method.

This paper addresses the tuning of data-driven controllers for poorly damped linear time-invariant systems in the internal model control (IMC) architecture. In this study, fictitious reference iterative tuning (FRIT), which is one of the controller parameter tuning methods with the data obtained from a one-shot experiment, is used for tuning the controller. The Kautz expansion method in which the coefficients are tunable parameters is introduced to approximate the dynamics of linear time-invariant systems, which have poor damping characteristics. Such an approximated model with tunable parameters is implemented in the IMC architecture. A model and a controller can be realized simultaneously with a one-shot experiment by tuning the IMC with the parameterized Kautz expansion model and by using FRIT. The validity of the proposed method is examined with a numerical example.

Proportional-integral-derivative (PID) control has been widely used in industrial equipment. However, the transient response obtained is poor when the fixed PID controller is used for nonlinear systems. Moreover, a hydraulic excavator consists of a nonlinear system with a derivative element. In this case, the system output does not track the reference signal because the integral element in the controller is canceled out by the derivative element in the system. In order to improve the transient response and the tracking performance, a data-driven PID control for tuning and updating PID gains has been proposed. However, the data-driven PID control method could not improve the tracking performance. In this paper, the controller design scheme based on a data-driven approach for the hydraulic excavator is proposed. Moreover, the control parameters of the proposed scheme are updated in an off-line manner by using fictitious reference iterative tuning. The effectiveness of this controller is verified by simulating the control behavior of a hydraulic excavator.

The purpose of this study is to realize a near-field touch interface that is compact, flexible, and highly accurate. We applied a 3-dimensional image sensor (time-of-flight camera) to achieve the basic functions of conventional touch interfaces, such as clicking, dragging, and sliding, and we designed a complete projector-sensor system. Unlike conventional touch interfaces, such as those on tablet PCs, the system can sense the 3-dimensional positions of fingertips and 3-dimensional directions of fingers. Moreover, it does not require a real touch screen but instead utilizes a mobile projector for display. Nonetheless, the system is compact, with a working distance of as short as around 30 cm. Our methods solve the shadow and reflection problems of the time-of-flight camera and can provide robust detection results. Tests have shown that our approach has a high success rate (98.4%) on touch/hover detection and a small standard error (2.21 mm) on position detection on average for different participants, which is the best performance we have achieved. Some applications, such as the virtual keyboard and virtual joystick, are also realized based on the proposed projector-sensor system.