Data-driven Control Research Articles

Optimal direct data-driven control with stability guarantees

For model-free optimal control design, this paper proposes an approach based on optimizing the reference model that is used in direct data-driven controller synthesis. Optimality is defined with respect to suitable cost functions reflecting desired performance and control objectives. We rely on the well-known Virtual Reference Feedback Tuning technique and on a direct control design approach that ensures stability of the resulting closed-loop system. The proposed design method leads to a non-convex optimization problem with a small number of variables that can be easily solved by a global optimizer, such as by particle swarm optimization. The effectiveness of the proposed solution is illustrated in simulation examples.

Data-driven adaptive predictive control for an activated sludge process

Data-driven control requires no information of the mathematical model of the controlled process. This paper proposes the direct identification of controller parameters of activated sludge process. This class of data-driven control calculates the predictive controller parameters directly using subspace identification technique. By updating input-output data using receding window mechanism, the adaptive strategy can be achieved. The robustness test and stability analysis of direct adaptive model predictive control are discussed to realize the effectiveness of this adaptive control scheme. The applicability of the controller algorithm to adapt into varying kinetic parameters and operating conditions is evaluated. Simulation results show that by a proper and effective excitation of direct identification of controller parameters, the convergence and stability of the implicit predictive model can be achieved.

Data-driven optimal PID type ILC for a class of nonlinear batch process

The paper presents model-free proportional–integral–derivative (PID) type iterative learning control (ILC) approach for the nonlinear batch process. The dynamic linearisation method is considered, which uses the input-output (I/O) measurements to update the model at each iteration. Based on the newly updated model and error information of the previous iteration, optimal PID gains are updated iteratively. The quadratic performance index is employed to optimise the parameters of the PID controller, and then an optimal PID type data-driven iterative learning control (DDILC) scheme is established for nonlinear batch process. The convergence analysis of optimal PID type DDILC is also discussed which can be enhanced by the proper choice of penalty matrices. Simulation examples are also given to demonstrate the effectiveness of the proposed scheme.

Data-based stabilization of unknown bilinear systems with guaranteed basin of attraction

Motivated by the goal of having a building block in the direct design of data-driven controllers for nonlinear systems, we show how, for an unknown discrete-time bilinear system, the data collected in an offline open-loop experiment enable us to design a feedback controller and provide a guaranteed under-approximation of its basin of attraction. Both can be obtained by solving a linear matrix inequality for a fixed scalar parameter, and possibly iterating on different values of that parameter. The results of this data-based approach are compared with the ideal case when the model is known perfectly.

An $${\mathscr {H}}_{\infty }$$ Approach to Data-Driven Offset-Free Tracking

Data-driven controllers also called model-free controllers were invented in order to omit plant modeling step of model-based controllers. Design procedure of these controllers is directly based on experimental I/O data collected from real plant. It can ensure their reliability in real world applications, where the exact model is not available in most cases. In this paper, we consider the problem of accurate tracking performance in presence of external disturbances using data-driven methodologies combined with $$\mathscr {H}_\infty $$ approach. Defining the improved subspace-based predictor, as the base step of the proposed controller’s design procedure, an integrator is applied to the control loop, which increases the accuracy of controller’s reference tracking performance. Moreover, a weighting function is considered for disturbance attenuation. Simulation results evidently illustrate efficiency and satisfactory performance of the proposed controller.

In-Field Performance Optimization for mm-Wave Mixed-Signal Doherty Power Amplifiers: A Bandit Approach

Advanced power amplifier (PA) architectures are critical for 5G communication which requires PAs to have sufficient output power, high linearity, and high energy efficiency. The time-variant operational environment further demands the self-reconfigurability in the PA design. Recent developments in bandit-problems and reinforcement learning (RL) have led to data-driven control algorithms. This paper presents various RL-based algorithms for Doherty PA control, which achieve robust adaptive operation over environmental changes. Multiple RL frameworks are incorporated in the control algorithms including multi-armed bandit (MAB), continuum-armed bandit (CAB), contextual-bandit (CB), and actor-critic with experience replay (AC). The control algorithms based on the latter three frameworks leverage on the prior information about the Doherty PA’s characteristics to improve learning efficiency. In our simulation test where the optimal policy needs to be adjusted due to transmitter output load impedance mismatch, the MAB-based control learns the optimal policy within 25,000 samples. The control algorithms based on CAB and CB learn the optimal policy within 5,000 samples. The fastest learning rate is achieved by the control algorithm based on AC, which learns the optimal policy within 1,500 samples.

Comprehensive evaluation of a data driven control strategy: Experimental application to a pharmaceutical crystallization process

In this contribution, a data-driven control approach was developed and applied experimentally to a pharmaceutical batch cooling crystallization process. In this approach, a radial basis functions (RBF) network model was trained in real-time with experimental data (time varied temperature and chord length distribution) with two different input data update strategies. The control objective was to optimize the cooling profile with the aid of trained RBF to achieve the desired crystal population profile throughout the process. The robustness of the proposed control strategy was tested with 10 comprehensive experiments in the presence of several disturbances (initial supersaturation, impeller speed, water composition and seed size). The presented control strategy was able to easily handle all the case scenarios. In 8 cases, the experimental crystal population profile followed successfully the reference with less than 10% offset. In the remaining 2 cases, the offset was 17% that was due to the absence of the supersaturation. The proposed RBF network-driven control is a promising strategy that is easy to implement, fully-automated and relies on relatively limited data for training. Therefore, the RBF control is expected to contribute to quick process development and control, especially when there is a lack of comprehensive process understanding and historical data especially in the pharmaceutical industry.

Grid-Supporting Battery Energy Storage Systems in Islanded Microgrids: A Data-Driven Control Approach

Islanded microgrids have low real and reactive power generation capacity and low inertia. This makes them susceptible to large frequency and voltage deviations, which deteriorate power quality and can cause frequency or voltage collapse. Grid-supporting battery energy storage systems are a possible solution as they are able to respond quickly to changes of their real and reactive power set-points. In this paper, a data-driven grid-supporting control system for battery energy storage systems, which requires no changes to the inverters inner real and reactive power control loops compared with a conventional grid-supporting inverter, is proposed. Tuning the data-driven controller does not require a dynamic model of the microgrid. Instead, the frequency response of the microgrid is identified and used directly to optimally tune the controller for H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> performance and robustness criteria. The performance of the data-driven controller is verified through real-time software-in-the-loop electromagnetic-transient simulation, where it is compared with an inverse-droop controller and is shown to significantly reduce voltage and frequency deviations.

Constrained robust model predictive control embedded with a new data‐driven technique

In the control field, the adaptive model predictive control (AMPC) has the capability of taking effective control actions on unknown-but-bounded time-independent or slowly time-varying systems coupling with constraints. In essence, AMPC estimates the uncertain parameters or uncertainty set online by utilising historian data to extract model information. The model estimation procedure imposes some specific conditions on data and these extra conditions have restricted its practical use. To overcome these problems, a new data-driven control methodology is presented that integrates the data-driven concept into robust model predictive control (RMPC) architecture for unknown-but-bounded time-independent or slowly time-varying plant. The key novelty is to employ historian data to derive control policy and make a prediction in replacement with the complicated procedure of utilising data to estimate model parameters. A data-driven RMPC algorithm is developed within the robust model predictive control framework with the fulfilment of recursive feasibility and stability. The authors display the highlights of the data-driven model predictive control controller through two simulation examples. The resulting controller is verified to reduce conservativeness and increase the closed-loop performance of the system.

Model‐free adaptive control method for a class of unknown MIMO systems with measurement noise and application to quadrotor aircraft

In this work, a novel robust model-free adaptive control (Ro-MFAC) algorithm for a class of unknown multiple-input multiple-output (MIMO) systems with measurement noise is presented. The proposed algorithm, designing by a dynamic linearisation data model with the concept of the pseudo-Jacobian matrix and an adaptive decreasing factor, is a pure a data-driven control method, and only the input-output data are involved for the control system design. The introduction of the adaptive decreasing factor is to attenuate the noise effect on the performance for improving the robustness of the algorithm. The stability of the Ro-MFAC proposed algorithm is proven by rigorous mathematical theory, and the effectiveness of the Ro-MFAC is verified by a series of numerical simulations. Furthermore, the Ro-MFAC is applied to an attitude adjustment problem of a practical quadrotor aircraft for demonstrating the applicability of the proposed approach.

New trends in modeling and control of hybrid systems

New trends in modeling and control of hybrid systems

Compensated model-free adaptive tracking control scheme for autonomous underwater vehicles via extended state observer

This paper investigates the problem of precise trajectory tracking control of an autonomous underwater vehicle (AUV) under parameter perturbations and external disturbances. To design a data-driven control scheme with high tracking accuracy and strong robustness, a compensated model-free adaptive control (MFAC) scheme is proposed by combining an extended state observer (ESO). Specifically, a data-driven structure-improved linear ESO (SLESO) is derived to online estimate the model approximation error generated by pseudo Jacobian matrix estimation in the typical MFAC scheme. Another problem of the typical MFAC is that it cannot be directly applied to robotic systems with rotation in the inertial frame. To tackle this issue, a two-loop controller architecture is used to design the proposed SLESO-MFAC scheme. In addition, the structure of the mathematical model of AUVs and force analysis are used in MFAC design for the first time, thus providing a straightforward initial value setting method with explicit physical interpretation and helping to judge if the assumptions in the MFAC scheme can be satisfied. Furthermore, the stability analysis of the designed control system is given. Finally, the robustness and effectiveness of the proposed SLESO-MFAC scheme are substantiated via simulations and comparisons using a realistic dynamic model of the Falcon AUV.

From Noisy Data to Feedback Controllers: Nonconservative Design via a Matrix S-Lemma

In this article, we propose a new method to obtain feedback controllers of an unknown dynamical system directly from noisy input/state data. The key ingredient of our design is a new matrix S-lemma that will be proven in this article. We provide both strict and nonstrict versions of this S-lemma, which are of interest in their own right. Thereafter, we will apply these results to data-driven control. In particular, we will derive nonconservative design methods for quadratic stabilization, <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathcal {H}_2$</tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\mathcal {H}_{\infty }$</tex-math></inline-formula> control, all in terms of data-based linear matrix inequalities. In contrast to previous work, the dimensions of our decision variables are independent of the time horizon of the experiment. Our approach, thus, enables control design from large datasets.

Model‐free output feedback discrete sliding mode control with disturbance compensation for precision motion systems

This study proposes a novel model-free output feedback discrete sliding mode control (OFDSMC) with disturbance compensation to achieve high performance for precision motion systems encountering model uncertainties and disturbances. Traditional discrete sliding mode control design requires model and/or state information. In this study, OFDSMC with disturbance compensation is first developed by using a parametric input–output model such that no system states are needed. The essential four control terms of OFDSMC with disturbance compensation is revealed. Then a data-driven control method is introduced into OFDSMC to eliminate the need for the parametric model. The proposed model-free OFDSMC facilitates a rapid implementation without either a parameter model or a state observer, and achieves good robustness against model uncertainties and disturbances. Finally, three simulation cases are presented to illustrate the effectiveness and enhanced performance of the proposed approach.

Hybrid data-driven fuzzy active disturbance rejection control for tower crane systems

This paper proposes the Virtual Reference Feedback Tuning (VRFT) of a combination of two control algorithms, Active Disturbance Rejection Control (ADRC) as a representative data-driven (or model-free) control algorithm and fuzzy control, in order to exploit the advantages of data-driven control and fuzzy control. The combination of Active Disturbance Rejection Control with Proportional-Derivative Takagi-Sugeno Fuzzy Control (PDTSFC) tuned by Virtual Reference Feedback Tuning results in two novel data-driven algorithms referred to as hybrid data-driven fuzzy ADRC algorithms. The main benefit of this combination is the automatic optimal tuning in a model-free manner of the parameters of the combination of Active Disturbance Rejection Control with Proportional-Derivative Takagi-Sugeno Fuzzy Control called ADRC-PDTSFC. The second benefit is that the suggested combination is time saving in finding the optimal parameters of the controllers. However, since Virtual Reference Feedback Tuning generally works with linear controllers to solve a certain optimization problem and the fuzzy controllers are essentially nonlinear, this paper replaces the least-squares algorithm specific to Virtual Reference Feedback Tuning with a metaheuristic optimization algorithm, i.e. Grey Wolf Optimizer. The fuzzy control system stability is guaranteed by including a limit cycle-based stability analysis approach in Grey Wolf Optimizer algorithm to validate the next solution candidates. The hybrid data-driven fuzzy ADRC algorithms are validated as controllers in terms of real-time experiments conducted on three-degree-of-freedom tower crane system laboratory equipment. To determine the efficiency of the new hybrid data-driven fuzzy ADRC algorithms, their performance is compared experimentally with that of two control algorithms, namely Active Disturbance Rejection Control with Proportional-Derivative Takagi-Sugeno Fuzzy Control, whose parameters are optimally tuned by Grey Wolf Optimizer in a model-based manner using the nonlinear process model.

Data-Driven Robust Optimization for Greenhouse Temperature Control Using Model Predictive Control

This work proposes a novel data-driven robust model predictive control (DDRMPC) framework for automatic control of greenhouse temperature and CO2 concentration level. The essential concept is to combine dynamic models of greenhouse temperature and CO2 concentration level with data-driven models that identify uncertainty in weather forecast error. By leveraging a machine learning approach, support vector clustering with weighted generalized intersection kernel, data-driven uncertainty sets for ambient temperature and solar radiation are constructed from historical weather data. A training-calibration procedure that tunes the size of uncertainty sets is implemented to ensure that data-driven uncertainty sets attain appropriate performance guarantee. In order to solve the optimization problem in DDRMPC, an affine disturbance feedback policy that provides tractable approximations of optimal control is utilized. A case study of controlling temperature and CO2 concentration level in a greenhouse is carried out. The results show that the proposed DDRMPC framework can prevent the greenhouse climate from becoming harmful to plant and fruit. DDRMPC approach ends up with 20 % less total economic cost than rule-based control strategy. The proposed DDRMPC approach also gives better control performance comparing to certainty equivalent MPC and robust MPC.

Future vehicles: learnable wheeled robots

As one of the important signs of the third wave of artificial intelligence, wheeled robots not only inherit knowledge but also learn independently, which brings about to learnable wheeled robots that use a driving brain to achieve data-driven control and learning. Presently, most existing technologies for self-driving vehicles can learn positively from the benchmark drivers to guarantee safe driving. However, in many unpredicted situations, such as rollover, human drivers often cause the behavior of irrational subconscious on account of human emotions like panic. In this paper, we propose a learnable wheeled robot using the driving brain by taking the rollover as an example, which is the most serious and dangerous situation in dynamic vehicle operations. Then, based on the analysis of rollover accidents, we utilize the driving brain reversely and conduct negative learning, materializing, and condensing the group intelligence of accident experts, to solve the problem of the lack of individual intelligence in emergencies and further promote real-time response to other dangerous conditions, such as puncture for self-driving vehicles.

Approximately Optimal Control of Discrete-Time Nonlinear Switched Systems Using Globalized Dual Heuristic Programming

Based on the idea of data-driven control, a novel iterative adaptive dynamic programming (ADP) algorithm based on the globalized dual heuristic programming (GDHP) technique is used to solve the optimal control problem of discrete-time nonlinear switched systems. In order to solve the Hamilton–Jacobi–Bellman (HJB) equation of switched systems, the iterative ADP method is proposed and the strict convergence analysis is also provided. Three neural networks are constructed to implement the iterative ADP algorithm, where a novel model network is designed to identify the system dynamics, a critic network is used to approximate the cost function and its partial derivatives, and an action network is provided to obtain the approximate optimal control law. Two simulation examples are described to illustrate the effectiveness of the proposed method by comparing with the heuristic dynamic programming (HDP) and dual heuristic programming (DHP) methods.

Distributed Feature Selection for Multi-Class Classification Using ADMM

Feature selection is an important task in data-driven control applications to identify relevant features and remove non-informative ones, for example residual selection for fault diagnosis. For multi-class data, the objective is to find a minimal set of features that can distinguish data from all different classes. A distributed feature selection algorithm is derived using convex optimization and the Alternating Direction Method of Multipliers. The distributed algorithm scales well with increasing number of classes by utilizing parallel computations. Two case studies are used to evaluate the developed feature selection algorithm: fault classification of an internal combustion engine and the MNIST data set to illustrate a larger multi-class classification problem.

Model-Free Dynamic Voltage Control of Distributed Energy Resource (DER)-Based Microgrids

In this paper, we present a new control technique for sustaining dynamic voltage stability by effective reactive power control and coordination of distributed energy resources (DERs) in microgrids. The proposed control technique is based on model-free control (MFC), which has shown successful operation and improved performance in different domains and applications. This paper presents its first use in the voltage stability of a microgrid setting employing multiple synchronous generator (SG)-based and power electronic (PE)-based DERs. MFC is a computationally efficient, data-driven control technique that does not require modelling of the different components and disturbances in the power system. It is utilized as an online controller to achieve the dynamic voltage stability of a microgrid system under different disturbances and fault conditions. A 21-bus microgrid system fed by multiple DERs is considered as a case study and the overall dynamic voltage stability is investigated using time-domain dynamic simulations. Numerical results show that the proposed MFC provides improvements on the dynamic load bus voltage profiles and requires less computational time as compared to the traditional enhanced microgrid voltage stabilizer (EMGVS) scheme. Due to its simplicity and low computational requirement, MFC can be easily implemented in resource-constrained computing devices such as smart inverters.