Online Self-tuning Research Articles

Design Decoupling Controller based on Feedforward and Improved PID for Tension System

It is difficult to precisely control the substrate tension due to the tension system of roll-to-roll coating machine with strong coupling and time-varying parameters. Therefore, a feedforward improved proportional integral derivative (PID) parameters self-tuning decoupling controller of tension system is proposed in this paper. According to the components and working principle of the coating machine, the nonlinear coupling mechanism model of the global tension system is established, and the model is linearized by using the Taylor series. Based on the linear model, a feedforward improved PID parameters self-tuning decoupling controller is designed for tension system. Through the feedforward control compensates the modeled tension and speed interference, a tracking differentiator (TD) is used to improve the performance of PID, and the online self-tuning of PID parameters can be realized by an Radial Basis Function (RBF) identification network. The simulation results show that the proposed decoupling control strategy of tension system can well suppress the coupling interference between the units, and realize the real-time tuning of PID parameters.

Observer-Based Asynchronous Communication and Adaptive Control Protocols With Intermittent Updating and Interaction

This paper focuses on the distributed observer-based synchronization control of multi-agent systems (MASs) over wireless networks. Unlike existing works mostly focusing on the communication reduction of control sides, a perception-based data reduction and transmission approach is proposed to design an asynchronous communication and control mechanism, which can avoid continuous inter-agents communications for both the communication as well as control sides. Then, an observer-based asynchronously distributed controller with intermittent updating and interaction based on the proposed perception-based data-reduction communication scheme. The criteria of new data transmission of communication triggering strategy and control triggering strategy are developed, respectively. Moreover, it is further extended to develop an observer-based asynchronous projection-type adaptive control strategy for online self-tuning of parameters. Theoretical analysis shows that the proposed two asynchronous communication and control strategies not only ensure that all agents stabilize to their initial average state without requiring any global information, but also ensure the reduction of both communications and controller update frequencies. An example is demonstrated to verify the theoretical analysis.

Reinforcement Learning Control of Hydraulic Servo System Based on TD3 Algorithm

This paper aims at the characteristics of nonlinear, time-varying and parameter coupling in a hydraulic servo system. An intelligent control method is designed that uses self-learning without a model or prior knowledge, in order to achieve certain control effects. The control quantity can be obtained at the current moment through the continuous iteration of a strategy–value network, and the online self-tuning of parameters can be realized. Taking the hydraulic servo system as the experimental object, a twin delayed deep deterministic (TD3) policy gradient was used to reinforce the learning of the system. Additionally, the parameter setting was compared using a deep deterministic policy gradient (DDPG) and a linear–quadratic–Gaussian (LQG) based on linear quadratic Gaussian objective function. To compile the reinforcement learning algorithm and deploy it to the test platform controller for testing, we used the Speedgoat prototype target machine as the controller to build the fast prototype control test platform. MATLAB/Coder and compute unified device architecture (CUDA) were used to generate an S-function. The results show that, compared with other parameter tuning methods, the proposed algorithm can effectively optimize the controller parameters and improve the dynamic response of the system when tracking signals.

Dynamic Deadband Event-Triggered Strategy for Distributed Adaptive Consensus Control With Applications to Circuit Systems

This paper focuses on the distributed consensus seeking of multi-agent systems (MASs) with discrete-time control updating and intermittent communications among agents. Compared with existing linearly coupled protocols, a nonlinear coupled Zeno-free event-triggered controller is first proposed, which is further to project the static and dynamic triggering mechanisms exploited by using the deadband control method. Then, the node-based nonlinear coupled adaptive event-triggered controller with online self-tuning of time-varying coupling weight and its corresponding to static and dynamic deadband-based event-triggered mechanisms are designed, respectively. The exploited adaptive event-triggered controller does not rely on any global information of interaction structure and is implemented in a fully distributed fashion. In addition, two dynamic proposals not only cover existing static strategies as special cases, but also show that the minimal inter-execution time of dynamic one is not smaller than that of static one. Theoretical analysis shows that the proposed static and dynamic deadband-based event-triggered mechanisms can not only ensure the average consensus with Zeno-freeness, but also achieve the data reduction of communication and control. Finally, the proposed algorithms applied to circuit implementation are corroborated to prove its practical merits and validity.

Optimal Control Strategy of Back-to-Back Converter Based on AC/DC Voltage Source Converter

The back-to-back converter is a converter system composed of two voltage source converters (VSC). Because VSC can regulate the output of active and reactive power, the back-to-back converter has a good dynamic and static performance. However, not all passive loads are three-phase symmetrical. The existing control strategy may lead to asymmetric output voltage when back-to-back converter is used to supply unbalance load. Usually, an inner loop d/q decoupling controller, a constant DC voltage controller of the rectifier side, and a constant AC voltage controller of the inverter side are established. In this article, in order to improve the rectifier side, fuzzy online self-tuning is used to increase PI parameters on the outer voltage loop. PI control is replaced by PR control on the inner current loop. The improved rectifier control can reduce the overshoot and the output of harmonics. For the inverter side, the three-phase voltage will be imbalanced. Therefore, using PR control, which can track a sinusoidal signal without steady-state error, the control strategy of inverter is changed into “outer loop constant voltage and PR control-inner loop current d/q decoupling control.” This strategy can solve the unbalance problem of three-phase AC output. The simulation results show that the control strategy can solve the problem of asymmetric output voltage of the inverter side effectively.

Adaptive predictive control method for improving control stability of air-conditioning terminal in public buildings

Due to the characteristic of large scale, multiple equipment, variable working conditions and complex debugging of air conditioning system in public buildings, the energy efficiency and operating level of the system are limited in the process of operation, and the actual operation effect will deviate from the design expectation. In addition, it is difficult to build an accurate mechanism model and to implement corresponding online control, so the commissioning for controller parameter tuning has to rely on human experience. Meanwhile, the response of indoor temperature to system regulation variables is different for different terminals due to lag characteristic, which would further affect the stability for the overall system. To solve the above problems, an adaptive predictive control method is proposed based on the physical-data driven hybrid model, which could not only realize online self-tuning of terminal controller but also improve control stability of terminal and overall system. Then, taking variable air volume air-conditioning system as the study object, the effectiveness of the proposed method is validated through comparative experiments. Results show that the proposed method can improve control stability of terminal and overall system. Achievements will have important theoretical significance and wide application prospect for improving the adaptability ability and the overall operating level of air-conditioning system in public buildings, which has complex and time-varying operation condition.

A novel deep reinforcement learning enabled sparsity promoting adaptive control method to improve the stability of power systems with wind energy penetration

With increasing proportion of wind energy in power systems, the intermittence of such energy makes the system run a wide range of operating conditions. In this context, ordinary power system stabilizers (PSS) tuned based on the linearized model of the system at one operating condition may not be able to effectively damp low frequency oscillations (LFO), which brings great challenges to the stability of the system. To this end, this paper proposes a novel sparsity promoting adaptive control method for the online self-tuning of the PSS parameter settings. Different from the existing adaptive control methods, the proposed method combines deep deterministic policy gradient (DDPG) algorithm and sensitivity analysis theory to train an agent to learn the sparse coordinated control policy of multi-PSS. After training, the well-trained agent can be employed for online sparse coordinated adaptive control, and the control signal is only applied, when it is required and only to the key PSS parameters that have the maximum influence on the system stability. Simulation results verify that the proposed method can make the PSS achieve the better performance of damping oscillation and robustness against the change of wind energy in comparison with other methods.

Application of Deep Reinforcement Learning to Thermal Control of Space Telescope

Abstract With the development of deep space exploration technology, thermal control systems for space telescopes are becoming increasingly complex, leading to the key parameters of conventional thermal control systems are difficult to adjust online automatically. To achieve these adjustments, this paper provided detailed verification of the application of deep reinforcement learning to space telescope thermal control from three perspectives: thermophysical modeling, intelligent sensing-based radiator, and online self-tuning of thermal control parameters. This paper presents a high-speed and high-precision thermophysical modeling strategy in matlab/simulink with better computational efficiency than conventional approaches. And an intelligent sensing-based radiator is proposed that can realize autonomous regulation of the radiating cold plate by sensing the external space environment and the thermal load inside the spacecraft. A strategy for online self-tuning of the thermal control parameters based on deep reinforcement learning is also proposed. Theoretical and experimental results show that deep reinforcement learning thermal control (DRLPID) can achieve temperature control accuracy of 0.05 °C. The steady-state errors in the simulations were reduced by 22.7%, 37.4%, and 47.4% when compared with the reinforcement learning proportional–integral–derivative (PID), the neural network PID, and the fuzzy PID, respectively. The experimental steady-state errors were reduced by 20.4%, 32.5%, and 42.7%, respectively.

Research on Key Technologies of Elders’ Exoskeleton Robot Assisted by Physical Exercise Based on Fuzzy PID Control

In the past, walking aids (such as crutches and walking aids) widely used by the elderly in physical exercise had relatively single functions, low intelligence, insufficient safety and comfort, and could not meet the needs of users for independent living. Therefore, it is necessary to apply robot technology to the field of traditional walking aid equipment and design a more intelligent, comfortable and safe walking aid robot system. The gait control of the exoskeleton robot assisted by physical exercise for the elderly is studied, and an algorithm based on fuzzy PID is proposed. Firstly, a fuzzy PID control algorithm for the exoskeleton robot assisted by physical exercise for the elderly is given. Secondly, membership functions are selected according to the parameters of the robot joint angle error and error variation. Finally, combined with practical operation experience, a fuzzy rule table is established, and online self-tuning of PID parameters is completed. The experimental results show that the fuzzy PID control method can make the exoskeleton robot move smoothly and realize the gait control of the robot initially.

Design and Adaptive Terminal Sliding Mode Control of a Fast Tool Servo System for Diamond Machining of Freeform Surfaces

This paper presents the methodology for the design and control of a high-stiffness fast tool servo (FTS) system for practical application in diamond machining of freeform surfaces. The electromechanical model of the FTS system is established by combining the piezoelectric actuator and the flexure hinge mechanism for the formulation of the control methodology. In order to accommodate the system nonlinearity, cutting force disturbances as well as to realize fast and accurate tracking of designated tool path required for diamond machining of freeform surfaces, a novel adaptive terminal sliding mode (NATS) control strategy capable of online self-tuning of the system gain and reach time related parameters are developed. Stability analysis of the proposed NATS controller is performed. The convergence behavior of tracking error in the presence of model parameter variations and cutting force disturbance is analyzed. Experimental verification was conducted to validate the effectiveness and robustness of the NATS control. By comparing with other related control methods, it shows significantly improved performance. The percentage of tracking error was only 0.1% for the varying trajectories on machining water-drop surface. Taking advantage of the excellent tracking performances of the FTS prototype on a diamond turning machine, various freeform surfaces were successfully produced with high quality.

Disturbance feedforward control for active vibration isolation systems with internal isolator dynamics

This paper presents a disturbance feedforward control strategy for active vibration isolation systems that have internal isolator dynamics. The structure of the feedforward controller follows from modeling, but to achieve robust performance the tuning of the controller parameters will be based on measurement data. More specifically, the controller poles, which correspond to internal isolator dynamics, are obtained offline from system identification experiments, while the controller zeros are obtained from online self-tuning using measured sampled-data systems. Self-tuning is used to reduce parameter uncertainty when having too limited identification possibilities, and to account automatically for the effect of noise amplification. Measurement results obtained from an industrial vibration isolation system show that residual vibrations are largely reduced up to the level of output noise limitations.

Power control based on particle swarm optimization of grid-connected inverter for hybrid renewable energy system

This paper is focused on the study of particle swarm optimization (PSO)-based PI controllers for the power control of a grid-connected inverter supplied from a hybrid renewable energy system. It is composed of two renewable energy sources (wind turbine and photovoltaic – PV – solar panels) and two energy storage systems (battery and hydrogen system, integrated by fuel cell and electrolyzer). Three PSO-based PI controllers are implemented: (1) conventional PI controller with offline tuning by PSO algorithm based on the integral time absolute error (ITAE) index; (2) PI controllers with online self-tuning by PSO algorithm based on the error; and (3) PI controllers with online self-tuning by PSO algorithm based on the ITAE index. To evaluate and compare the three controllers, the hybrid renewable energy system and the grid-connected inverter are simulated under changes in the active and reactive power values, as well as under a grid voltage sag. The results show that the online PSO-based PI controllers that optimize the ITAE index achieves the best response.

Real-time on line tuning of fuzzy controller for two-link rigid–flexible robot manipulators

In this paper, an online time self-tuning multi-input–multi-output (MIMO) fuzzy bang-bang controller (FBBC) is proposed for the control of two-link rigid and flexible robot manipulators. Two-link rigid and flexible robot manipulators are highly non-linear plants. The fuzzy control is based on the Takagi–Sugeno-type architecture fuzzy model combined with online self-tuning so that both the desired transient and steady-state responses can be achieved. The proposed FBBC is different from a fuzzy logic controller (FLC) in that it has a bi-level output like a relay but with fuzzy inputs. The online self-tuning is based on the gradient of steepest descent tuning method, which tunes the FBBC’s input and output gains. The controller operation is demonstrated and compared with a classic FLC and sliding mode controller (SMC) by simulation to highlight its tracking ability and the manipulator’s positioning control with rigid and flexible robot types. Based on the simulation results, the proposed controller with this tuning strategy was found to be superior at different operating conditions.

Nonlinear Internal Model Control Using Echo State Network for Pneumatic Muscle System

Pneumatic muscle (PM) has many advantages such as light weight, high power to weight ratio and low price. However, it has strong time varying characteristic. The complex nonlinear dynamics of PM system poses some challenges for achieving accurate modeling and control. To solve these problems, we propose nonlinear internal model control (IMC) using echo state network (ESN) for PM system in this paper. The ESN based IMC is termed ESNBIMC, which fully embodies the virtues of ESN and IMC. In ESNBIMC, the dynamic model of PM system is identified by an ESN. The other ESN is trained to learn the inverse dynamics of the system, and then it can be used as a nonlinear controller. Recursive Least Square (RLS) algorithm can be applied to online training the ESN without affecting the previous weight structure, which is very suitable for real-time control problems. By using the identification ability of ESN and RLS, high accurate plant model of PM system without detailed model information can be built. In addition, strong robustness also can be attained by online self-tuning of controller and internal model. Experiment demonstrates the effectiveness of the proposed control algorithm. The results show that ESNBIMC achieves satisfactory tracking performance for PM system.

Tuning of PID Parameters and Fuzzy Adaptive PID Control of the Hydrostatic Driving System

Composed of a variable displacement pump and a constant displacement motor, the hydrostatic driving system is a kind of closed speed control system with adjustable displacement. It is widely used in the field of engineering vehicle and other fields. Based on an analysis of the constitution and mathematical model of the hydrostatic driving system, the present study tuned PID parameters by using the critical proportioning method and the optimization method of NCD respectively. Then a kind of fuzzy adaptive PID controller was designed on the basis of the traditional PID control and the fuzzy control theory. In the controller, fuzzy logic was used to realize online self-tuning of PID parameters according to the motor speed error and its derivative, so that the system could have better adaptive ability and strong disturbance resisting performance. The dynamic simulation was made in MATLAB/SIMULINK. The simulation results show that the optimization method of NCD has better tuning effect and the response performance of the fuzzy adaptive PID controller is better than that of the classic one. Besides, it should be noted that a drawback was found about the fuzzy adaptive PID control. On the basis of fixed scale factors, a group of quantification factors is appropriate for a specific input signal, but for other signals, the response of the system is not so ideal. A method of adjusting quantification factors according to input signal was adopted to solve the above problem. Automatic adjusting of quantification factors was realized, and this could ensure ideal response to all input signals.

Human Simulated Intelligent Controller with Fuzzy Online Self-Tuning of Parameters and its Application to a Cart-Double Pendulum

Using the basic concepts and design methods of Human Simulated Intelligent Control (HSIC), we have designed the master control level of an HSIC controller for the swinging-up and handstand control of a cart-double pendulum. Then, the self-tuning structure of the self-tuning level of the HSIC Controller is implemented using fuzzy logic rules. This structure has fuzzy self-tuning abilities in the swing-up control of a cart-double pendulum system and achieves online self-tuning of the control parameters of the HSIC controller efficiently with a hierarchical and multimode control structure. The computer simulation and realtime experiments of the swing-up control of a cart-double pendulum system show that the fuzzy online self-tuning of the control parameters markedly enhances the robustness and adaptability of the HSIC.

Development of a GPC-based PID controller for unstable systems with deadtime

Till now, traditional low-order control schemes have never been applicable to unstable systems with deadtime. In this paper, we present the first application of a PID controller with time-scheduled gains to unstable systems with deadtime, consisting of a single unstable pole. The control gains are designed based on a generalised predictive control (GPC) approach. The only user specifications required are simple and classical desired properties as in the natural frequency and the damping ratio of the closed-loop system. An approach is further developed to subsequent on-line self-tuning of control weights so that the overall control system remains applicable and effective in the face of disturbances and slowly varying dynamics. A detailed analysis of the closed-loop stability of the thus designed control system is further provided in the paper. Based on stability conditions developed, the prediction horizon for the GPC-based controller may be effectively computed. Finally, simulation examples illustrate the performance of the control system.